publication_date
stringlengths 20
20
| doc_url
stringlengths 33
34
| id
stringlengths 33
34
| authors
stringlengths 5
20.1k
| update_date
stringlengths 20
20
| category_all
stringlengths 5
138
| abstract
stringlengths 123
3.09k
| category_primary
stringclasses 78
values | title
stringlengths 10
211
|
|---|---|---|---|---|---|---|---|---|
2024-06-14T17:38:21Z | http://arxiv.org/pdf/2406.10203v1 | http://arxiv.org/abs/2406.10203v1 | Naaman Tan, Josef Valvoda, Anej Svete, Tianyu Liu, Yanxia Qin, Kan Min-Yen, Ryan Cotterell | 2024-06-14T17:38:21Z | cs.CL | The relationship between the quality of a string and its probability
$p(\boldsymbol{y})$ under a language model has been influential in the
development of techniques to build good text generation systems. For example,
several decoding algorithms have been motivated to manipulate
$p(\boldsymbol{y})$ to produce higher-quality text. In this work, we examine
the probability--quality relationship in language models explicitly aligned to
human preferences, e.g., through Reinforcement Learning through Human Feedback
(RLHF). We find that, given a general language model and its aligned version,
for corpora sampled from an aligned language model, there exists a trade-off
between the average reward and average log-likelihood of the strings under the
general language model. We provide a formal treatment of this issue and
demonstrate how a choice of sampling adaptor allows for a selection of how much
likelihood we exchange for the reward. | cs.CL | A Fundamental Trade-off in Aligned Language Models and its Relation to Sampling Adaptors |
2024-06-14T17:23:57Z | http://arxiv.org/pdf/2406.10190v2 | http://arxiv.org/abs/2406.10190v2 | Alexander Gurung, Mirella Lapata | 2024-06-26T14:22:18Z | cs.CL | Characters are integral to long-form narratives, but are poorly understood by
existing story analysis and generation systems. While prior work has simplified
characters via graph-based methods and brief character descriptions, we aim to
better tackle the problem of representing complex characters by taking
inspiration from advice given to professional writers. We propose CHIRON, a new
`character sheet' based representation that organizes and filters textual
information about characters. We construct CHIRON sheets in two steps: a
Generation Module that prompts an LLM for character information via
question-answering and a Validation Module that uses automated reasoning and a
domain-specific entailment model to eliminate false facts about a character. We
validate CHIRON via the downstream task of masked-character prediction, where
our experiments show CHIRON is better and more flexible than comparable
summary-based baselines. We also show that metrics derived from CHIRON can be
used to automatically infer character-centricity in stories, and that these
metrics align with human judgments. | cs.CL | CHIRON: Rich Character Representations in Long-Form Narratives |
2024-06-14T16:56:40Z | http://arxiv.org/pdf/2406.10177v1 | http://arxiv.org/abs/2406.10177v1 | Dena Mujtaba, Nihar R. Mahapatra, Megan Arney, J. Scott Yaruss, Caryn Herring, Jia Bin | 2024-06-14T16:56:40Z | eess.AS, cs.CL, I.2; K.4 | Automatic speech recognition (ASR) systems often falter while processing
stuttering-related disfluencies -- such as involuntary blocks and word
repetitions -- yielding inaccurate transcripts. A critical barrier to progress
is the scarcity of large, annotated disfluent speech datasets. Therefore, we
present an inclusive ASR design approach, leveraging large-scale
self-supervised learning on standard speech followed by targeted fine-tuning
and data augmentation on a smaller, curated dataset of disfluent speech. Our
data augmentation technique enriches training datasets with various
disfluencies, enhancing ASR processing of these speech patterns. Results show
that fine-tuning wav2vec 2.0 with even a relatively small, labeled dataset,
alongside data augmentation, can significantly reduce word error rates for
disfluent speech. Our approach not only advances ASR inclusivity for people who
stutter, but also paves the way for ASRs that can accommodate wider speech
variations. | eess.AS | Inclusive ASR for Disfluent Speech: Cascaded Large-Scale Self-Supervised Learning with Targeted Fine-Tuning and Data Augmentation |
2024-06-14T16:54:48Z | http://arxiv.org/pdf/2406.10174v1 | http://arxiv.org/abs/2406.10174v1 | Mohamad Elzohbi, Richard Zhao | 2024-06-14T16:54:48Z | cs.CL | The intersection between poetry and music provides an interesting case for
computational creativity, yet remains relatively unexplored. This paper
explores the integration of poetry and music through the lens of beat patterns,
investigating whether a byte-based language model can generate words that fit
specific beat patterns within the context of poetry. Drawing on earlier
studies, we developed a method to train a byte-based transformer model, ByT5,
to align poems with beat patterns. The results demonstrate a high level of beat
alignment while maintaining semantic coherence. Future work will aim to improve
the model's ability to create complete beat-aligned poems. | cs.CL | Let the Poem Hit the Rhythm: Using a Byte-Based Transformer for Beat-Aligned Poetry Generation |
2024-06-14T16:51:21Z | http://arxiv.org/pdf/2406.10173v1 | http://arxiv.org/abs/2406.10173v1 | Wenxuan Ding, Weiqi Wang, Sze Heng Douglas Kwok, Minghao Liu, Tianqing Fang, Jiaxin Bai, Junxian He, Yangqiu Song | 2024-06-14T16:51:21Z | cs.CL | Enhancing Language Models' (LMs) ability to understand purchase intentions in
E-commerce scenarios is crucial for their effective assistance in various
downstream tasks. However, previous approaches that distill intentions from LMs
often fail to generate meaningful and human-centric intentions applicable in
real-world E-commerce contexts. This raises concerns about the true
comprehension and utilization of purchase intentions by LMs. In this paper, we
present IntentionQA, a double-task multiple-choice question answering benchmark
to evaluate LMs' comprehension of purchase intentions in E-commerce.
Specifically, LMs are tasked to infer intentions based on purchased products
and utilize them to predict additional purchases. IntentionQA consists of 4,360
carefully curated problems across three difficulty levels, constructed using an
automated pipeline to ensure scalability on large E-commerce platforms. Human
evaluations demonstrate the high quality and low false-negative rate of our
benchmark. Extensive experiments across 19 language models show that they still
struggle with certain scenarios, such as understanding products and intentions
accurately, jointly reasoning with products and intentions, and more, in which
they fall far behind human performances. Our code and data are publicly
available at https://github.com/HKUST-KnowComp/IntentionQA. | cs.CL | IntentionQA: A Benchmark for Evaluating Purchase Intention Comprehension Abilities of Language Models in E-commerce |
2024-06-14T16:50:29Z | http://arxiv.org/pdf/2406.10172v1 | http://arxiv.org/abs/2406.10172v1 | Matteo Gabburo, Stefano Campese, Federico Agostini, Alessandro Moschitti | 2024-06-14T16:50:29Z | cs.CL | Answer Sentence Selection (AS2) is a critical task for designing effective
retrieval-based Question Answering (QA) systems. Most advancements in AS2 focus
on English due to the scarcity of annotated datasets for other languages. This
lack of resources prevents the training of effective AS2 models in different
languages, creating a performance gap between QA systems in English and other
locales. In this paper, we introduce new high-quality datasets for AS2 in five
European languages (French, German, Italian, Portuguese, and Spanish), obtained
through supervised Automatic Machine Translation (AMT) of existing English AS2
datasets such as ASNQ, WikiQA, and TREC-QA using a Large Language Model (LLM).
We evaluated our approach and the quality of the translated datasets through
multiple experiments with different Transformer architectures. The results
indicate that our datasets are pivotal in producing robust and powerful
multilingual AS2 models, significantly contributing to closing the performance
gap between English and other languages. | cs.CL | Datasets for Multilingual Answer Sentence Selection |
2024-06-14T16:26:20Z | http://arxiv.org/pdf/2406.10162v3 | http://arxiv.org/abs/2406.10162v3 | Carson Denison, Monte MacDiarmid, Fazl Barez, David Duvenaud, Shauna Kravec, Samuel Marks, Nicholas Schiefer, Ryan Soklaski, Alex Tamkin, Jared Kaplan, Buck Shlegeris, Samuel R. Bowman, Ethan Perez, Evan Hubinger | 2024-06-29T00:28:47Z | cs.AI, cs.CL | In reinforcement learning, specification gaming occurs when AI systems learn
undesired behaviors that are highly rewarded due to misspecified training
goals. Specification gaming can range from simple behaviors like sycophancy to
sophisticated and pernicious behaviors like reward-tampering, where a model
directly modifies its own reward mechanism. However, these more pernicious
behaviors may be too complex to be discovered via exploration. In this paper,
we study whether Large Language Model (LLM) assistants which find easily
discovered forms of specification gaming will generalize to perform rarer and
more blatant forms, up to and including reward-tampering. We construct a
curriculum of increasingly sophisticated gameable environments and find that
training on early-curriculum environments leads to more specification gaming on
remaining environments. Strikingly, a small but non-negligible proportion of
the time, LLM assistants trained on the full curriculum generalize zero-shot to
directly rewriting their own reward function. Retraining an LLM not to game
early-curriculum environments mitigates, but does not eliminate,
reward-tampering in later environments. Moreover, adding harmlessness training
to our gameable environments does not prevent reward-tampering. These results
demonstrate that LLMs can generalize from common forms of specification gaming
to more pernicious reward tampering and that such behavior may be nontrivial to
remove. | cs.AI | Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models |
2024-06-14T16:00:29Z | http://arxiv.org/pdf/2406.10149v1 | http://arxiv.org/abs/2406.10149v1 | Yuri Kuratov, Aydar Bulatov, Petr Anokhin, Ivan Rodkin, Dmitry Sorokin, Artyom Sorokin, Mikhail Burtsev | 2024-06-14T16:00:29Z | cs.CL, cs.AI | In recent years, the input context sizes of large language models (LLMs) have
increased dramatically. However, existing evaluation methods have not kept
pace, failing to comprehensively assess the efficiency of models in handling
long contexts. To bridge this gap, we introduce the BABILong benchmark,
designed to test language models' ability to reason across facts distributed in
extremely long documents. BABILong includes a diverse set of 20 reasoning
tasks, including fact chaining, simple induction, deduction, counting, and
handling lists/sets. These tasks are challenging on their own, and even more
demanding when the required facts are scattered across long natural text. Our
evaluations show that popular LLMs effectively utilize only 10-20\% of the
context and their performance declines sharply with increased reasoning
complexity. Among alternatives to in-context reasoning, Retrieval-Augmented
Generation methods achieve a modest 60\% accuracy on single-fact question
answering, independent of context length. Among context extension methods, the
highest performance is demonstrated by recurrent memory transformers, enabling
the processing of lengths up to 11 million tokens. The BABILong benchmark is
extendable to any length to support the evaluation of new upcoming models with
increased capabilities, and we provide splits up to 1 million token lengths. | cs.CL | BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystack |
2024-06-14T15:42:42Z | http://arxiv.org/pdf/2406.10133v1 | http://arxiv.org/abs/2406.10133v1 | Smilla Due, Sneha Das, Marianne Andersen, Berta Plandolit López, Sniff Andersen Nexø, Line Clemmensen | 2024-06-14T15:42:42Z | cs.CL, cs.AI | Large Language Models (LLMs) have an increasing impact on our lives with use
cases such as chatbots, study support, coding support, ideation, writing
assistance, and more. Previous studies have revealed linguistic biases in
pronouns used to describe professions or adjectives used to describe men vs
women. These issues have to some degree been addressed in updated LLM versions,
at least to pass existing tests. However, biases may still be present in the
models, and repeated use of gender stereotypical language may reinforce the
underlying assumptions and are therefore important to examine further. This
paper investigates gender biases in LLMs in relation to educational choices
through an open-ended, true to user-case experimental design and a quantitative
analysis. We investigate the biases in the context of four different cultures,
languages, and educational systems (English/US/UK, Danish/DK, Catalan/ES, and
Hindi/IN) for ages ranging from 10 to 16 years, corresponding to important
educational transition points in the different countries. We find that there
are significant and large differences in the ratio of STEM to non-STEM
suggested education paths provided by chatGPT when using typical girl vs boy
names to prompt lists of suggested things to become. There are generally fewer
STEM suggestions in the Danish, Spanish, and Indian context compared to the
English. We also find subtle differences in the suggested professions, which we
categorise and report. | cs.CL | Evaluation of Large Language Models: STEM education and Gender Stereotypes |
2024-06-14T15:41:06Z | http://arxiv.org/pdf/2406.10130v1 | http://arxiv.org/abs/2406.10130v1 | Yan Liu, Yu Liu, Xiaokang Chen, Pin-Yu Chen, Daoguang Zan, Min-Yen Kan, Tsung-Yi Ho | 2024-06-14T15:41:06Z | cs.CL | Pre-trained Language models (PLMs) have been acknowledged to contain harmful
information, such as social biases, which may cause negative social impacts or
even bring catastrophic results in application. Previous works on this problem
mainly focused on using black-box methods such as probing to detect and
quantify social biases in PLMs by observing model outputs. As a result,
previous debiasing methods mainly finetune or even pre-train language models on
newly constructed anti-stereotypical datasets, which are high-cost. In this
work, we try to unveil the mystery of social bias inside language models by
introducing the concept of {\sc Social Bias Neurons}. Specifically, we propose
{\sc Integrated Gap Gradients (IG$^2$)} to accurately pinpoint units (i.e.,
neurons) in a language model that can be attributed to undesirable behavior,
such as social bias. By formalizing undesirable behavior as a distributional
property of language, we employ sentiment-bearing prompts to elicit classes of
sensitive words (demographics) correlated with such sentiments. Our IG$^2$ thus
attributes the uneven distribution for different demographics to specific
Social Bias Neurons, which track the trail of unwanted behavior inside PLM
units to achieve interoperability. Moreover, derived from our interpretable
technique, {\sc Bias Neuron Suppression (BNS)} is further proposed to mitigate
social biases. By studying BERT, RoBERTa, and their attributable differences
from debiased FairBERTa, IG$^2$ allows us to locate and suppress identified
neurons, and further mitigate undesired behaviors. As measured by prior metrics
from StereoSet, our model achieves a higher degree of fairness while
maintaining language modeling ability with low cost. | cs.CL | The Devil is in the Neurons: Interpreting and Mitigating Social Biases in Pre-trained Language Models |
2024-06-14T15:23:39Z | http://arxiv.org/pdf/2406.10118v1 | http://arxiv.org/abs/2406.10118v1 | Holy Lovenia, Rahmad Mahendra, Salsabil Maulana Akbar, Lester James V. Miranda, Jennifer Santoso, Elyanah Aco, Akhdan Fadhilah, Jonibek Mansurov, Joseph Marvin Imperial, Onno P. Kampman, Joel Ruben Antony Moniz, Muhammad Ravi Shulthan Habibi, Frederikus Hudi, Railey Montalan, Ryan Ignatius, Joanito Agili Lopo, William Nixon, Börje F. Karlsson, James Jaya, Ryandito Diandaru, Yuze Gao, Patrick Amadeus, Bin Wang, Jan Christian Blaise Cruz, Chenxi Whitehouse, Ivan Halim Parmonangan, Maria Khelli, Wenyu Zhang, Lucky Susanto, Reynard Adha Ryanda, Sonny Lazuardi Hermawan, Dan John Velasco, Muhammad Dehan Al Kautsar, Willy Fitra Hendria, Yasmin Moslem, Noah Flynn, Muhammad Farid Adilazuarda, Haochen Li, Johanes Lee, R. Damanhuri, Shuo Sun, Muhammad Reza Qorib, Amirbek Djanibekov, Wei Qi Leong, Quyet V. Do, Niklas Muennighoff, Tanrada Pansuwan, Ilham Firdausi Putra, Yan Xu, Ngee Chia Tai, Ayu Purwarianti, Sebastian Ruder, William Tjhi, Peerat Limkonchotiwat, Alham Fikri Aji, Sedrick Keh, Genta Indra Winata, Ruochen Zhang, Fajri Koto, Zheng-Xin Yong, Samuel Cahyawijaya | 2024-06-14T15:23:39Z | cs.CL | Southeast Asia (SEA) is a region rich in linguistic diversity and cultural
variety, with over 1,300 indigenous languages and a population of 671 million
people. However, prevailing AI models suffer from a significant lack of
representation of texts, images, and audio datasets from SEA, compromising the
quality of AI models for SEA languages. Evaluating models for SEA languages is
challenging due to the scarcity of high-quality datasets, compounded by the
dominance of English training data, raising concerns about potential cultural
misrepresentation. To address these challenges, we introduce SEACrowd, a
collaborative initiative that consolidates a comprehensive resource hub that
fills the resource gap by providing standardized corpora in nearly 1,000 SEA
languages across three modalities. Through our SEACrowd benchmarks, we assess
the quality of AI models on 36 indigenous languages across 13 tasks, offering
valuable insights into the current AI landscape in SEA. Furthermore, we propose
strategies to facilitate greater AI advancements, maximizing potential utility
and resource equity for the future of AI in SEA. | cs.CL | SEACrowd: A Multilingual Multimodal Data Hub and Benchmark Suite for Southeast Asian Languages |
2024-06-14T14:56:04Z | http://arxiv.org/pdf/2406.10099v1 | http://arxiv.org/abs/2406.10099v1 | Jiaqi Li, Yixuan Tang, Yi Yang | 2024-06-14T14:56:04Z | cs.CL | Large language models (LLMs) have demonstrated remarkable capabilities across
various tasks but still face challenges such as hallucinations. One potential
reason for hallucinations is the lack of relevant knowledge or context. Thus, a
promising solution to mitigate this issue involves instructing LLMs to respond
with "I do not know" when a question falls outside their knowledge domain or
the provided context. However, in this work, we observed that LLMs struggle to
admit their lack of knowledge, primarily due to existing instruction datasets
designed to encourage specific answers. To improve large language models'
capability to recognize the boundaries of their knowledge, we propose a novel
approach called uncertainty-sensitive tuning. This method involves two-stage
training designed for uncertainty recognition and prompt-sensitive activation.
In the first stage, we guide the LLM to reject unknown questions. In the second
stage, we recover the decreased performance in QA tasks by incorporating
designed causal instructions. By leveraging this method, we aim to enhance the
model's ability to identify areas of uncertainty. The experimental results
demonstrate that our proposed uncertainty-sensitive tuning method significantly
improves the performance of the Llama2-chat-7B model. Specifically, it achieves
a substantial 34.7% improvement in handling questions involving knowledge gaps
compared to the original model. Moreover, our approach outperforms GPT-4,
exhibiting a 9.4% increase in overall performance. We open-source the model and
code on GitHub. | cs.CL | Know the Unknown: An Uncertainty-Sensitive Method for LLM Instruction Tuning |
2024-06-14T14:47:19Z | http://arxiv.org/pdf/2406.10091v1 | http://arxiv.org/abs/2406.10091v1 | Xiaoman Wang, Claudio Fantinuoli | 2024-06-14T14:47:19Z | cs.CL | Assessing the performance of interpreting services is a complex task, given
the nuanced nature of spoken language translation, the strategies that
interpreters apply, and the diverse expectations of users. The complexity of
this task become even more pronounced when automated evaluation methods are
applied. This is particularly true because interpreted texts exhibit less
linearity between the source and target languages due to the strategies
employed by the interpreter.
This study aims to assess the reliability of automatic metrics in evaluating
simultaneous interpretations by analyzing their correlation with human
evaluations. We focus on a particular feature of interpretation quality, namely
translation accuracy or faithfulness. As a benchmark we use human assessments
performed by language experts, and evaluate how well sentence embeddings and
Large Language Models correlate with them. We quantify semantic similarity
between the source and translated texts without relying on a reference
translation. The results suggest GPT models, particularly GPT-3.5 with direct
prompting, demonstrate the strongest correlation with human judgment in terms
of semantic similarity between source and target texts, even when evaluating
short textual segments. Additionally, the study reveals that the size of the
context window has a notable impact on this correlation. | cs.CL | Exploring the Correlation between Human and Machine Evaluation of Simultaneous Speech Translation |
2024-06-14T14:41:44Z | http://arxiv.org/pdf/2406.10086v2 | http://arxiv.org/abs/2406.10086v2 | Megan Ayers, Luke Sanford, Margaret Roberts, Eddie Yang | 2024-06-21T18:14:42Z | cs.CL, cs.LG, stat.ME | Experimental methods for estimating the impacts of text on human evaluation
have been widely used in the social sciences. However, researchers in
experimental settings are usually limited to testing a small number of
pre-specified text treatments. While efforts to mine unstructured texts for
features that causally affect outcomes have been ongoing in recent years, these
models have primarily focused on the topics or specific words of text, which
may not always be the mechanism of the effect. We connect these efforts with
NLP interpretability techniques and present a method for flexibly discovering
clusters of similar text phrases that are predictive of human reactions to
texts using convolutional neural networks. When used in an experimental
setting, this method can identify text treatments and their effects under
certain assumptions. We apply the method to two datasets. The first enables
direct validation of the model's ability to detect phrases known to cause the
outcome. The second demonstrates its ability to flexibly discover text
treatments with varying textual structures. In both cases, the model learns a
greater variety of text treatments compared to benchmark methods, and these
text features quantitatively meet or exceed the ability of benchmark methods to
predict the outcome. | cs.CL | Discovering influential text using convolutional neural networks |
2024-06-14T14:40:10Z | http://arxiv.org/pdf/2406.10085v1 | http://arxiv.org/abs/2406.10085v1 | Ashim Gupta, Vivek Gupta, Shuo Zhang, Yujie He, Ning Zhang, Shalin Shah | 2024-06-14T14:40:10Z | cs.CL | To completely understand a document, the use of textual information is not
enough. Understanding visual cues, such as layouts and charts, is also
required. While the current state-of-the-art approaches for document
understanding (both OCR-based and OCR-free) work well, a thorough analysis of
their capabilities and limitations has not yet been performed. Therefore, in
this work, we addresses the limitation of current VisualQA models when applied
to charts and plots. To investigate shortcomings of the state-of-the-art
models, we conduct a comprehensive behavioral analysis, using ChartQA as a case
study. Our findings indicate that existing models particularly underperform in
answering questions related to the chart's structural and visual context, as
well as numerical information. To address these issues, we propose three simple
pre-training tasks that enforce the existing model in terms of both
structural-visual knowledge, as well as its understanding of numerical
questions. We evaluate our pre-trained model (called MatCha-v2) on three chart
datasets - both extractive and abstractive question datasets - and observe that
it achieves an average improvement of 1.7% over the baseline model. | cs.CL | Enhancing Question Answering on Charts Through Effective Pre-training Tasks |
2024-06-14T14:37:52Z | http://arxiv.org/pdf/2406.10083v1 | http://arxiv.org/abs/2406.10083v1 | Siddhant Arora, Ankita Pasad, Chung-Ming Chien, Jionghao Han, Roshan Sharma, Jee-weon Jung, Hira Dhamyal, William Chen, Suwon Shon, Hung-yi Lee, Karen Livescu, Shinji Watanabe | 2024-06-14T14:37:52Z | cs.CL, cs.SD, eess.AS | The Spoken Language Understanding Evaluation (SLUE) suite of benchmark tasks
was recently introduced to address the need for open resources and benchmarking
of complex spoken language understanding (SLU) tasks, including both
classification and sequence generation tasks, on natural speech. The benchmark
has demonstrated preliminary success in using pre-trained speech foundation
models (SFM) for these SLU tasks. However, the community still lacks a
fine-grained understanding of the comparative utility of different SFMs.
Inspired by this, we ask: which SFMs offer the most benefits for these complex
SLU tasks, and what is the most effective approach for incorporating these
SFMs? To answer this, we perform an extensive evaluation of multiple supervised
and self-supervised SFMs using several evaluation protocols: (i) frozen SFMs
with a lightweight prediction head, (ii) frozen SFMs with a complex prediction
head, and (iii) fine-tuned SFMs with a lightweight prediction head. Although
the supervised SFMs are pre-trained on much more speech recognition data (with
labels), they do not always outperform self-supervised SFMs; the latter tend to
perform at least as well as, and sometimes better than, supervised SFMs,
especially on the sequence generation tasks in SLUE. While there is no
universally optimal way of incorporating SFMs, the complex prediction head
gives the best performance for most tasks, although it increases the inference
time. We also introduce an open-source toolkit and performance leaderboard,
SLUE-PERB, for these tasks and modeling strategies. | cs.CL | On the Evaluation of Speech Foundation Models for Spoken Language Understanding |
2024-06-14T14:28:06Z | http://arxiv.org/pdf/2406.10073v1 | http://arxiv.org/abs/2406.10073v1 | Rémi Uro, Marie Tahon, David Doukhan, Antoine Laurent, Albert Rilliard | 2024-06-14T14:28:06Z | eess.AS, cs.CL, cs.HC, cs.SD | Transition Relevance Places are defined as the end of an utterance where the
interlocutor may take the floor without interrupting the current speaker
--i.e., a place where the turn is terminal. Analyzing turn terminality is
useful to study the dynamic of turn-taking in spontaneous conversations. This
paper presents an automatic classification of spoken utterances as Terminal or
Non-Terminal in multi-speaker settings. We compared audio, text, and fusions of
both approaches on a French corpus of TV and Radio extracts annotated with
turn-terminality information at each speaker change. Our models are based on
pre-trained self-supervised representations. We report results for different
fusion strategies and varying context sizes. This study also questions the
problem of performance variability by analyzing the differences in results for
multiple training runs with random initialization. The measured accuracy would
allow the use of these models for large-scale analysis of turn-taking. | eess.AS | Detecting the terminality of speech-turn boundary for spoken interactions in French TV and Radio content |
2024-06-14T14:24:02Z | http://arxiv.org/pdf/2406.14522v1 | http://arxiv.org/abs/2406.14522v1 | Mikhail V. Tamm, Els Heinsalu, Stefano Scialla, Marco Patriarca | 2024-06-14T14:24:02Z | physics.soc-ph, cs.CL | We introduce a language competition model that incorporates the effects of
memory and learning on the language shift dynamics, using the Abrams-Strogatz
model as a starting point. On a coarse grained time scale, the effects of
memory and learning can be expressed as thresholds on the speakers fractions.
In its simplest form, the resulting model is exactly solvable. Besides the
consensus on one of the two languages, the model describes additional
equilibrium states that are not present in the Abrams-Strogatz model: a stable
coexistence of the two languages, if both thresholds are low enough, so that
the language shift processes in the two opposite directions compensate each
other, and a frozen state coinciding with the initial state, when both
thresholds are too high for any language shift to take place. We show
numerically that these results are preserved for threshold functions of a more
general shape. | physics.soc-ph | Learning thresholds lead to stable language coexistence |
2024-06-14T14:07:26Z | http://arxiv.org/pdf/2406.10052v1 | http://arxiv.org/abs/2406.10052v1 | Haoyu Wang, Guoqiang Hu, Guodong Lin, Wei-Qiang Zhang, Jian Li | 2024-06-14T14:07:26Z | cs.SD, cs.CL, eess.AS | As a robust and large-scale multilingual speech recognition model, Whisper
has demonstrated impressive results in many low-resource and
out-of-distribution scenarios. However, its encoder-decoder structure hinders
its application to streaming speech recognition. In this paper, we introduce
Simul-Whisper, which uses the time alignment embedded in Whisper's
cross-attention to guide auto-regressive decoding and achieve chunk-based
streaming ASR without any fine-tuning of the pre-trained model. Furthermore, we
observe the negative effect of the truncated words at the chunk boundaries on
the decoding results and propose an integrate-and-fire-based truncation
detection model to address this issue. Experiments on multiple languages and
Whisper architectures show that Simul-Whisper achieves an average absolute word
error rate degradation of only 1.46% at a chunk size of 1 second, which
significantly outperforms the current state-of-the-art baseline. | cs.SD | Simul-Whisper: Attention-Guided Streaming Whisper with Truncation Detection |
2024-06-14T13:54:29Z | http://arxiv.org/pdf/2406.12925v1 | http://arxiv.org/abs/2406.12925v1 | Ihor Stepanov, Mykhailo Shtopko | 2024-06-14T13:54:29Z | cs.LG, cs.AI, cs.CL, cs.IR | Information extraction tasks require both accurate, efficient, and
generalisable models. Classical supervised deep learning approaches can achieve
the required performance, but they need large datasets and are limited in their
ability to adapt to different tasks. On the other hand, large language models
(LLMs) demonstrate good generalization, meaning that they can adapt to many
different tasks based on user requests. However, LLMs are computationally
expensive and tend to fail to generate structured outputs. In this article, we
will introduce a new kind of GLiNER model that can be used for various
information extraction tasks while being a small encoder model. Our model
achieved SoTA performance on zero-shot NER benchmarks and leading performance
on question-answering, summarization and relation extraction tasks.
Additionally, in this article, we will cover experimental results on
self-learning approaches for named entity recognition using GLiNER models. | cs.LG | GLiNER multi-task: Generalist Lightweight Model for Various Information Extraction Tasks |
2024-06-14T13:49:07Z | http://arxiv.org/pdf/2406.10040v1 | http://arxiv.org/abs/2406.10040v1 | Jin Liu, Steffen Thoma | 2024-06-14T13:49:07Z | cs.CL, cs.AI | This paper describes the inference system of FZI-WIM at the SemEval-2024 Task
2: Safe Biomedical Natural Language Inference for Clinical Trials. Our system
utilizes the chain of thought (CoT) paradigm to tackle this complex reasoning
problem and further improves the CoT performance with self-consistency. Instead
of greedy decoding, we sample multiple reasoning chains with the same prompt
and make the final verification with majority voting. The self-consistent CoT
system achieves a baseline F1 score of 0.80 (1st), faithfulness score of 0.90
(3rd), and consistency score of 0.73 (12th). We release the code and data
publicly https://github.com/jens5588/FZI-WIM-NLI4CT. | cs.CL | FZI-WIM at SemEval-2024 Task 2: Self-Consistent CoT for Complex NLI in Biomedical Domain |
2024-06-14T13:32:43Z | http://arxiv.org/pdf/2406.10023v1 | http://arxiv.org/abs/2406.10023v1 | Luckeciano C. Melo, Panagiotis Tigas, Alessandro Abate, Yarin Gal | 2024-06-14T13:32:43Z | cs.LG, cs.CL, stat.ML | Leveraging human preferences for steering the behavior of Large Language
Models (LLMs) has demonstrated notable success in recent years. Nonetheless,
data selection and labeling are still a bottleneck for these systems,
particularly at large scale. Hence, selecting the most informative points for
acquiring human feedback may considerably reduce the cost of preference
labeling and unleash the further development of LLMs. Bayesian Active Learning
provides a principled framework for addressing this challenge and has
demonstrated remarkable success in diverse settings. However, previous attempts
to employ it for Preference Modeling did not meet such expectations. In this
work, we identify that naive epistemic uncertainty estimation leads to the
acquisition of redundant samples. We address this by proposing the Bayesian
Active Learner for Preference Modeling (BAL-PM), a novel stochastic acquisition
policy that not only targets points of high epistemic uncertainty according to
the preference model but also seeks to maximize the entropy of the acquired
prompt distribution in the feature space spanned by the employed LLM. Notably,
our experiments demonstrate that BAL-PM requires 33% to 68% fewer preference
labels in two popular human preference datasets and exceeds previous stochastic
Bayesian acquisition policies. | cs.LG | Deep Bayesian Active Learning for Preference Modeling in Large Language Models |
2024-06-14T13:29:36Z | http://arxiv.org/pdf/2406.10019v1 | http://arxiv.org/abs/2406.10019v1 | Mikhail Gorbunov, Nikolay Yudin, Vera Soboleva, Aibek Alanov, Alexey Naumov, Maxim Rakhuba | 2024-06-14T13:29:36Z | cs.LG, cs.AI, cs.CL, cs.CV, cs.NA, math.NA | The increasing size of neural networks has led to a growing demand for
methods of efficient fine-tuning. Recently, an orthogonal fine-tuning paradigm
was introduced that uses orthogonal matrices for adapting the weights of a
pretrained model. In this paper, we introduce a new class of structured
matrices, which unifies and generalizes structured classes from previous works.
We examine properties of this class and build a structured orthogonal
parametrization upon it. We then use this parametrization to modify the
orthogonal fine-tuning framework, improving parameter and computational
efficiency. We empirically validate our method on different domains, including
adapting of text-to-image diffusion models and downstream task fine-tuning in
language modeling. Additionally, we adapt our construction for orthogonal
convolutions and conduct experiments with 1-Lipschitz neural networks. | cs.LG | Group and Shuffle: Efficient Structured Orthogonal Parametrization |
2024-06-14T13:07:46Z | http://arxiv.org/pdf/2406.09994v1 | http://arxiv.org/abs/2406.09994v1 | Manas Jhalani, Annervaz K M, Pushpak Bhattacharyya | 2024-06-14T13:07:46Z | cs.CL | In the realm of multimodal tasks, Visual Question Answering (VQA) plays a
crucial role by addressing natural language questions grounded in visual
content. Knowledge-Based Visual Question Answering (KBVQA) advances this
concept by adding external knowledge along with images to respond to questions.
We introduce an approach for KBVQA, augmenting the existing vision-language
transformer encoder-decoder (OFA) model. Our main contribution involves
enhancing questions by incorporating relevant external knowledge extracted from
knowledge graphs, using a dynamic triple extraction method. We supply a
flexible number of triples from the knowledge graph as context, tailored to
meet the requirements for answering the question. Our model, enriched with
knowledge, demonstrates an average improvement of 4.75\% in Exact Match Score
over the state-of-the-art on three different KBVQA datasets. Through
experiments and analysis, we demonstrate that furnishing variable triples for
each question improves the reasoning capabilities of the language model in
contrast to supplying a fixed number of triples. This is illustrated even for
recent large language models. Additionally, we highlight the model's
generalization capability by showcasing its SOTA-beating performance on a small
dataset, achieved through straightforward fine-tuning. | cs.CL | Precision Empowers, Excess Distracts: Visual Question Answering With Dynamically Infused Knowledge In Language Models |
2024-06-14T12:52:42Z | http://arxiv.org/pdf/2406.09988v1 | http://arxiv.org/abs/2406.09988v1 | Xiaowen Sun, Xufeng Zhao, Jae Hee Lee, Wenhao Lu, Matthias Kerzel, Stefan Wermter | 2024-06-14T12:52:42Z | cs.AI, cs.CL, cs.RO | The state of an object reflects its current status or condition and is
important for a robot's task planning and manipulation. However, detecting an
object's state and generating a state-sensitive plan for robots is challenging.
Recently, pre-trained Large Language Models (LLMs) and Vision-Language Models
(VLMs) have shown impressive capabilities in generating plans. However, to the
best of our knowledge, there is hardly any investigation on whether LLMs or
VLMs can also generate object state-sensitive plans. To study this, we
introduce an Object State-Sensitive Agent (OSSA), a task-planning agent
empowered by pre-trained neural networks. We propose two methods for OSSA: (i)
a modular model consisting of a pre-trained vision processing module (dense
captioning model, DCM) and a natural language processing model (LLM), and (ii)
a monolithic model consisting only of a VLM. To quantitatively evaluate the
performances of the two methods, we use tabletop scenarios where the task is to
clear the table. We contribute a multimodal benchmark dataset that takes object
states into consideration. Our results show that both methods can be used for
object state-sensitive tasks, but the monolithic approach outperforms the
modular approach. The code for OSSA is available at
\url{https://github.com/Xiao-wen-Sun/OSSA} | cs.AI | Details Make a Difference: Object State-Sensitive Neurorobotic Task Planning |
2024-06-14T12:49:38Z | http://arxiv.org/pdf/2406.10313v1 | http://arxiv.org/abs/2406.10313v1 | Chen Chen, Zehua Liu, Xiaolou Li, Lantian Li, Dong Wang | 2024-06-14T12:49:38Z | cs.CL, cs.CV | The first Chinese Continuous Visual Speech Recognition Challenge aimed to
probe the performance of Large Vocabulary Continuous Visual Speech Recognition
(LVC-VSR) on two tasks: (1) Single-speaker VSR for a particular speaker and (2)
Multi-speaker VSR for a set of registered speakers. The challenge yielded
highly successful results, with the best submission significantly outperforming
the baseline, particularly in the single-speaker task. This paper
comprehensively reviews the challenge, encompassing the data profile, task
specifications, and baseline system construction. It also summarises the
representative techniques employed by the submitted systems, highlighting the
most effective approaches. Additional information and resources about this
challenge can be accessed through the official website at
http://cnceleb.org/competition. | cs.CL | CNVSRC 2023: The First Chinese Continuous Visual Speech Recognition Challenge |
2024-06-14T12:41:07Z | http://arxiv.org/pdf/2406.09979v1 | http://arxiv.org/abs/2406.09979v1 | Krish Goel, Mahek Chandak | 2024-06-14T12:41:07Z | cs.CL, cs.AI, cs.IR | Large Language Models (LLMs) excel in natural language tasks but face
limitations due to static training datasets, resulting in outdated or
contextually shallow responses. Retrieval-Augmented Generation (RAG) addresses
this by integrating real-time external knowledge, enhancing model accuracy and
credibility, especially for knowledge-intensive tasks. However, RAG-enhanced
LLMs struggle with long contexts, causing them to "choke" on information
overload, compromising response quality. Recent RAG applications use
hierarchical data structures for storing documents, organized at various levels
of summarization and information density. In this context, we introduce HIRO
(Hierarchical Information Retrieval Optimization), a novel querying approach
for RAG applications using hierarchical structures for storing documents. HIRO
employs DFS-based recursive similarity score calculation and branch pruning to
minimize the context returned to the LLM without informational loss. HIRO
outperforms existing querying mechanisms on the NarrativeQA dataset by an
absolute performance gain of 10.85%. | cs.CL | HIRO: Hierarchical Information Retrieval Optimization |
2024-06-14T12:39:39Z | http://arxiv.org/pdf/2406.09977v1 | http://arxiv.org/abs/2406.09977v1 | Maximilian Spliethöver, Sai Nikhil Menon, Henning Wachsmuth | 2024-06-14T12:39:39Z | cs.CL | Dialects introduce syntactic and lexical variations in language that occur in
regional or social groups. Most NLP methods are not sensitive to such
variations. This may lead to unfair behavior of the methods, conveying negative
bias towards dialect speakers. While previous work has studied dialect-related
fairness for aspects like hate speech, other aspects of biased language, such
as lewdness, remain fully unexplored. To fill this gap, we investigate
performance disparities between dialects in the detection of five aspects of
biased language and how to mitigate them. To alleviate bias, we present a
multitask learning approach that models dialect language as an auxiliary task
to incorporate syntactic and lexical variations. In our experiments with
African-American English dialect, we provide empirical evidence that
complementing common learning approaches with dialect modeling improves their
fairness. Furthermore, the results suggest that multitask learning achieves
state-of-the-art performance and helps to detect properties of biased language
more reliably. | cs.CL | Disentangling Dialect from Social Bias via Multitask Learning to Improve Fairness |
2024-06-14T12:38:49Z | http://arxiv.org/pdf/2406.10312v1 | http://arxiv.org/abs/2406.10312v1 | Yang Hu | 2024-06-14T12:38:49Z | cs.CL | Recall initiator identification and assessment are the preliminary steps to
prevent medical device recall. Conventional analysis tools are inappropriate
for processing massive and multi-formatted data comprehensively and completely
to meet the higher expectations of delicacy management with the increasing
overall data volume and textual data format. This study presents a
bigdata-analytics-based machine learning-natural language processing work tool
to address the shortcomings in dealing efficiency and data process versatility
of conventional tools in the practical context of big data volume and muti data
format. This study identified, assessed and analysed the medical device recall
initiators according to the public medical device recall database from 2018 to
2024 with the ML-NLP tool. The results suggest that the unsupervised
Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering
algorithm can present each single recall initiator in a specific manner,
therefore helping practitioners to identify the recall reasons comprehensively
and completely within a short time frame. This is then followed by text
similarity-based textual classification to assist practitioners in controlling
the group size of recall initiators and provide managerial insights from the
operational to the tactical and strategical levels. This ML-NLP work tool can
not only capture specific details of each recall initiator but also interpret
the inner connection of each existing initiator and can be implemented for risk
identification and assessment in the forward SC. Finally, this paper suggests
some concluding remarks and presents future works. More proactive practices and
control solutions for medical device recalls are expected in the future. | cs.CL | In-depth analysis of recall initiators of medical devices with a Machine Learning-Natural language Processing workflow |
2024-06-14T12:31:44Z | http://arxiv.org/pdf/2406.09972v1 | http://arxiv.org/abs/2406.09972v1 | KuanChao Chu, Yi-Pei Chen, Hideki Nakayama | 2024-06-14T12:31:44Z | cs.CL | This research investigates prompt designs of evaluating generated texts using
large language models (LLMs). While LLMs are increasingly used for scoring
various inputs, creating effective prompts for open-ended text evaluation
remains challenging due to model sensitivity and subjectivity in evaluation of
text generation. Our study experimented with different prompt structures,
altering the sequence of output instructions and including explanatory reasons.
We found that the order of presenting reasons and scores significantly
influences LLMs' scoring, with a different level of rule understanding in the
prompt. An additional optimization may enhance scoring alignment if sufficient
data is available. This insight is crucial for improving the accuracy and
consistency of LLM-based evaluations. | cs.CL | A Better LLM Evaluator for Text Generation: The Impact of Prompt Output Sequencing and Optimization |
2024-06-14T12:16:08Z | http://arxiv.org/pdf/2406.09967v1 | http://arxiv.org/abs/2406.09967v1 | Ine Gevers, Walter Daelemans | 2024-06-14T12:16:08Z | cs.CL | This study aims to acquire more insights into the continuous pre-training
phase of BERT regarding entity knowledge, using the COVID-19 pandemic as a case
study. Since the pandemic emerged after the last update of BERT's pre-training
data, the model has little to no entity knowledge about COVID-19. Using
continuous pre-training, we control what entity knowledge is available to the
model. We compare the baseline BERT model with the further pre-trained variants
on the fact-checking benchmark Check-COVID. To test the robustness of
continuous pre-training, we experiment with several adversarial methods to
manipulate the input data, such as training on misinformation and shuffling the
word order until the input becomes nonsensical. Surprisingly, our findings
reveal that these methods do not degrade, and sometimes even improve, the
model's downstream performance. This suggests that continuous pre-training of
BERT is robust against misinformation. Furthermore, we are releasing a new
dataset, consisting of original texts from academic publications in the
LitCovid repository and their AI-generated false counterparts. | cs.CL | Bag of Lies: Robustness in Continuous Pre-training BERT |
2024-06-14T12:10:51Z | http://arxiv.org/pdf/2406.09961v1 | http://arxiv.org/abs/2406.09961v1 | Chufan Shi, Cheng Yang, Yaxin Liu, Bo Shui, Junjie Wang, Mohan Jing, Linran Xu, Xinyu Zhu, Siheng Li, Yuxiang Zhang, Gongye Liu, Xiaomei Nie, Deng Cai, Yujiu Yang | 2024-06-14T12:10:51Z | cs.SE, cs.CL, cs.CV | We introduce a new benchmark, ChartMimic, aimed at assessing the
visually-grounded code generation capabilities of large multimodal models
(LMMs). ChartMimic utilizes information-intensive visual charts and textual
instructions as inputs, requiring LMMs to generate the corresponding code for
chart rendering. ChartMimic includes 1,000 human-curated (figure, instruction,
code) triplets, which represent the authentic chart use cases found in
scientific papers across various domains(e.g., Physics, Computer Science,
Economics, etc). These charts span 18 regular types and 4 advanced types,
diversifying into 191 subcategories. Furthermore, we propose multi-level
evaluation metrics to provide an automatic and thorough assessment of the
output code and the rendered charts. Unlike existing code generation
benchmarks, ChartMimic places emphasis on evaluating LMMs' capacity to
harmonize a blend of cognitive capabilities, encompassing visual understanding,
code generation, and cross-modal reasoning. The evaluation of 3 proprietary
models and 11 open-weight models highlights the substantial challenges posed by
ChartMimic. Even the advanced GPT-4V, Claude-3-opus only achieve an average
score of 73.2 and 53.7, respectively, indicating significant room for
improvement. We anticipate that ChartMimic will inspire the development of
LMMs, advancing the pursuit of artificial general intelligence. | cs.SE | ChartMimic: Evaluating LMM's Cross-Modal Reasoning Capability via Chart-to-Code Generation |
2024-06-14T11:58:49Z | http://arxiv.org/pdf/2406.09952v1 | http://arxiv.org/abs/2406.09952v1 | Imanol Miranda, Ander Salaberria, Eneko Agirre, Gorka Azkune | 2024-06-14T11:58:49Z | cs.CV, cs.CL, cs.LG | Existing Vision-Language Compositionality (VLC) benchmarks like SugarCrepe
are formulated as image-to-text retrieval problems, where, given an image, the
models need to select between the correct textual description and a synthetic
hard negative text. In this work we present the Bidirectional Vision-Language
Compositionality (BiVLC) dataset. The novelty of BiVLC is to add a synthetic
hard negative image generated from the synthetic text, resulting in two
image-to-text retrieval examples (one for each image) and, more importantly,
two text-to-image retrieval examples (one for each text). Human annotators
filter out ill-formed examples ensuring the validity of the benchmark. The
experiments on BiVLC uncover a weakness of current multimodal models, as they
perform poorly in the text-to-image direction. In fact, when considering both
retrieval directions, the conclusions obtained in previous works change
significantly. In addition to the benchmark, we show that a contrastive model
trained using synthetic images and texts improves the state of the art in
SugarCrepe and in BiVLC for both retrieval directions. The gap to human
performance in BiVLC confirms that Vision-Language Compositionality is still a
challenging problem. BiVLC and code are available at
https://imirandam.github.io/BiVLC_project_page. | cs.CV | BiVLC: Extending Vision-Language Compositionality Evaluation with Text-to-Image Retrieval |
2024-06-14T11:53:14Z | http://arxiv.org/pdf/2406.09950v1 | http://arxiv.org/abs/2406.09950v1 | Naijun Zheng, Xucheng Wan, Kai Liu, Ziqing Du, Zhou Huan | 2024-06-14T11:53:14Z | cs.SD, cs.CL, eess.AS | Although contextualized automatic speech recognition (ASR) systems are
commonly used to improve the recognition of uncommon words, their effectiveness
is hindered by the inherent limitations of speech-text data availability. To
address this challenge, our study proposes to leverage extensive text-only
datasets and contextualize pre-trained ASR models using a straightforward
text-augmentation (TA) technique, all while keeping computational costs
minimal. In particular, to contextualize a pre-trained CIF-based ASR, we
construct a codebook using limited speech-text data. By utilizing a simple
codebook lookup process, we convert available text-only data into latent text
embeddings. These embeddings then enhance the inputs for the contextualized
ASR. Our experiments on diverse Mandarin test sets demonstrate that our TA
approach significantly boosts recognition performance. The top-performing
system shows relative CER improvements of up to 30% on rare words and 15%
across all words in general. | cs.SD | An efficient text augmentation approach for contextualized Mandarin speech recognition |
2024-06-14T11:48:54Z | http://arxiv.org/pdf/2406.09948v1 | http://arxiv.org/abs/2406.09948v1 | Junho Myung, Nayeon Lee, Yi Zhou, Jiho Jin, Rifki Afina Putri, Dimosthenis Antypas, Hsuvas Borkakoty, Eunsu Kim, Carla Perez-Almendros, Abinew Ali Ayele, Víctor Gutiérrez-Basulto, Yazmín Ibáñez-García, Hwaran Lee, Shamsuddeen Hassan Muhammad, Kiwoong Park, Anar Sabuhi Rzayev, Nina White, Seid Muhie Yimam, Mohammad Taher Pilehvar, Nedjma Ousidhoum, Jose Camacho-Collados, Alice Oh | 2024-06-14T11:48:54Z | cs.CL | Large language models (LLMs) often lack culture-specific knowledge of daily
life, especially across diverse regions and non-English languages. Existing
benchmarks for evaluating LLMs' cultural sensitivities are limited to a single
language or collected from online sources such as Wikipedia, which do not
reflect the mundane everyday lifestyles of diverse regions. That is,
information about the food people eat for their birthday celebrations, spices
they typically use, musical instruments youngsters play, or the sports they
practice in school is common cultural knowledge but uncommon in easily
collected online sources, especially for underrepresented cultures. To address
this issue, we introduce BLEnD, a hand-crafted benchmark designed to evaluate
LLMs' everyday knowledge across diverse cultures and languages. BLEnD comprises
52.6k question-answer pairs from 16 countries/regions, in 13 different
languages, including low-resource ones such as Amharic, Assamese, Azerbaijani,
Hausa, and Sundanese. We construct the benchmark to include two formats of
questions: short-answer and multiple-choice. We show that LLMs perform better
for cultures that are highly represented online, with a maximum 57.34%
difference in GPT-4, the best-performing model, in the short-answer format. For
cultures represented by mid-to-high-resource languages, LLMs perform better in
their local languages, but for cultures represented by low-resource languages,
LLMs perform better in English than the local languages. We make our dataset
publicly available at: https://github.com/nlee0212/BLEnD. | cs.CL | BLEnD: A Benchmark for LLMs on Everyday Knowledge in Diverse Cultures and Languages |
2024-06-14T11:34:36Z | http://arxiv.org/pdf/2406.09938v1 | http://arxiv.org/abs/2406.09938v1 | Tim Menzner, Jochen L. Leidner | 2024-06-14T11:34:36Z | cs.CL, cs.AI | The World Wide Web provides unrivalled access to information globally,
including factual news reporting and commentary. However, state actors and
commercial players increasingly spread biased (distorted) or fake (non-factual)
information to promote their agendas. We compare several large, pre-trained
language models on the task of sentence-level news bias detection and sub-type
classification, providing quantitative and qualitative results. | cs.CL | Experiments in News Bias Detection with Pre-Trained Neural Transformers |
2024-06-14T11:10:17Z | http://arxiv.org/pdf/2406.09923v1 | http://arxiv.org/abs/2406.09923v1 | Mingyu Derek Ma, Chenchen Ye, Yu Yan, Xiaoxuan Wang, Peipei Ping, Timothy S Chang, Wei Wang | 2024-06-14T11:10:17Z | cs.CL, cs.AI, cs.LG | The integration of Artificial Intelligence (AI), especially Large Language
Models (LLMs), into the clinical diagnosis process offers significant potential
to improve the efficiency and accessibility of medical care. While LLMs have
shown some promise in the medical domain, their application in clinical
diagnosis remains underexplored, especially in real-world clinical practice,
where highly sophisticated, patient-specific decisions need to be made. Current
evaluations of LLMs in this field are often narrow in scope, focusing on
specific diseases or specialties and employing simplified diagnostic tasks. To
bridge this gap, we introduce CliBench, a novel benchmark developed from the
MIMIC IV dataset, offering a comprehensive and realistic assessment of LLMs'
capabilities in clinical diagnosis. This benchmark not only covers diagnoses
from a diverse range of medical cases across various specialties but also
incorporates tasks of clinical significance: treatment procedure
identification, lab test ordering and medication prescriptions. Supported by
structured output ontologies, CliBench enables a precise and multi-granular
evaluation, offering an in-depth understanding of LLM's capability on diverse
clinical tasks of desired granularity. We conduct a zero-shot evaluation of
leading LLMs to assess their proficiency in clinical decision-making. Our
preliminary results shed light on the potential and limitations of current LLMs
in clinical settings, providing valuable insights for future advancements in
LLM-powered healthcare. | cs.CL | CliBench: Multifaceted Evaluation of Large Language Models in Clinical Decisions on Diagnoses, Procedures, Lab Tests Orders and Prescriptions |
2024-06-14T11:02:21Z | http://arxiv.org/pdf/2406.09920v1 | http://arxiv.org/abs/2406.09920v1 | Amit Rozner, Barak Battash, Lior Wolf, Ofir Lindenbaum | 2024-06-14T11:02:21Z | cs.CL, cs.AI | Large Language Models (LLMs) can become outdated over time as they may lack
updated world knowledge, leading to factual knowledge errors and gaps.
Knowledge Editing (KE) aims to overcome this challenge using weight updates
that do not require expensive retraining. We propose treating KE as an LLM
alignment problem. Toward this goal, we introduce Knowledge Direct Preference
Optimization (KDPO), a variation of the Direct Preference Optimization (DPO)
that is more effective for knowledge modifications. Our method is based on an
online approach that continually updates the knowledge stored in the model. We
use the current knowledge as a negative sample and the new knowledge we want to
introduce as a positive sample in a process called DPO. We also use
teacher-forcing for negative sample generation and optimize using the positive
sample, which helps maintain localized changes. We tested our KE method on
various datasets and models, comparing it to several cutting-edge methods, with
100 and 500 sequential edits. Additionally, we conducted an ablation study
comparing our method to the standard DPO approach. Our experimental results
show that our modified DPO method allows for more refined KE, achieving similar
or better performance compared to previous methods. | cs.CL | Knowledge Editing in Language Models via Adapted Direct Preference Optimization |
2024-06-14T10:15:49Z | http://arxiv.org/pdf/2406.09900v1 | http://arxiv.org/abs/2406.09900v1 | Jie Wu, Yufeng Zhu, Lei Shen, Xuqing Lu | 2024-06-14T10:15:49Z | cs.CL | Recently developed large language models (LLMs) such as ChatGPT, Claude, and
Llama have demonstrated impressive abilities, and even surpass human-level
performance in several tasks. Despite their success, the resource-intensive
demands of these models, requiring significant computational power for both
training and inference, limit their deployment to high-performance servers.
Additionally, the extensive calculation requirements of the models often lead
to increased latency in response times. With the increasing need for LLMs to
operate efficiently on CPUs, research about lightweight models that are
optimized for CPU inference has emerged. In this work, we introduce GEB-1.3B, a
lightweight LLM trained on 550 billion tokens in both Chinese and English
languages. We employ novel training techniques, including ROPE,
Group-Query-Attention, and FlashAttention-2, to accelerate training while
maintaining model performance. Additionally, we fine-tune the model using 10
million samples of instruction data to enhance alignment. GEB-1.3B exhibits
outstanding performance on general benchmarks such as MMLU, C-Eval, and CMMLU,
outperforming comparative models such as MindLLM-1.3B and TinyLLaMA-1.1B.
Notably, the FP32 version of GEB-1.3B achieves commendable inference times on
CPUs, with ongoing efforts to further enhance speed through advanced
quantization techniques. The release of GEB-1.3B as an open-source model marks
a significant contribution to the development of lightweight LLMs, promising to
foster further research and innovation in the field. | cs.CL | GEB-1.3B: Open Lightweight Large Language Model |
2024-06-14T10:13:37Z | http://arxiv.org/pdf/2406.09897v1 | http://arxiv.org/abs/2406.09897v1 | Xindian Ma, Wenyuan Liu, Peng Zhang, Nan Xu | 2024-06-14T10:13:37Z | cs.CL | Inspired by the Bloch Sphere representation, we propose a novel rotary
position encoding on a three-dimensional sphere, named 3D Rotary Position
Encoding (3D-RPE). 3D-RPE is an advanced version of the widely used 2D Rotary
Position Encoding (RoPE), with two major advantages for modeling long contexts:
controllable long-term decay and improved position resolution. For controllable
long-term decay, 3D-RPE allows for the regulation of long-term decay within the
chunk size, ensuring the modeling of relative positional information between
tokens at a distant relative position. For enhanced position resolution, 3D-RPE
can mitigate the degradation of position resolution caused by position
interpolation on RoPE. We have conducted experiments on long-context Natural
Language Understanding (NLU) and long-sequence Language Modeling (LM) tasks.
From the experimental results, 3D-RPE achieved performance improvements over
RoPE, especially in long-context NLU tasks. | cs.CL | 3D-RPE: Enhancing Long-Context Modeling Through 3D Rotary Position Encoding |
2024-06-14T09:55:54Z | http://arxiv.org/pdf/2406.09884v1 | http://arxiv.org/abs/2406.09884v1 | Wanqing Zhao, Yuta Nakashima, Haiyuan Chen, Noboru Babaguchi | 2024-06-14T09:55:54Z | cs.MM, cs.CL, cs.SI | Fake news detection in social media has become increasingly important due to
the rapid proliferation of personal media channels and the consequential
dissemination of misleading information. Existing methods, which primarily rely
on multimodal features and graph-based techniques, have shown promising
performance in detecting fake news. However, they still face a limitation,
i.e., sparsity in graph connections, which hinders capturing possible
interactions among tweets. This challenge has motivated us to explore a novel
method that densifies the graph's connectivity to capture denser interaction
better. Our method constructs a cross-modal tweet graph using CLIP, which
encodes images and text into a unified space, allowing us to extract potential
connections based on similarities in text and images. We then design a Feature
Contextualization Network with Label Propagation (FCN-LP) to model the
interaction among tweets as well as positive or negative correlations between
predicted labels of connected tweets. The propagated labels from the graph are
weighted and aggregated for the final detection. To enhance the model's
generalization ability to unseen events, we introduce a domain generalization
loss that ensures consistent features between tweets on seen and unseen events.
We use three publicly available fake news datasets, Twitter, PHEME, and Weibo,
for evaluation. Our method consistently improves the performance over the
state-of-the-art methods on all benchmark datasets and effectively demonstrates
its aptitude for generalizing fake news detection in social media. | cs.MM | Enhancing Fake News Detection in Social Media via Label Propagation on Cross-modal Tweet Graph |
2024-06-14T09:52:27Z | http://arxiv.org/pdf/2406.09881v2 | http://arxiv.org/abs/2406.09881v2 | Yongkang Liu, Ercong Nie, Shi Feng, Zheng Hua, Zifeng Ding, Daling Wang, Yifei Zhang, Hinrich Schütze | 2024-06-28T12:58:11Z | cs.CL | Current state-of-the-art dialogue systems heavily rely on extensive training
datasets. However, challenges arise in domains where domain-specific training
datasets are insufficient or entirely absent. To tackle this challenge, we
propose a novel data \textbf{A}ugmentation framework for
\textbf{M}ulti-\textbf{D}omain \textbf{D}ialogue \textbf{G}eneration, referred
to as \textbf{AMD$^2$G}. The AMD$^2$G framework consists of a data augmentation
process and a two-stage training approach: domain-agnostic training and domain
adaptation training. We posit that domain corpora are a blend of
domain-agnostic and domain-specific features, with certain representation
patterns shared among diverse domains. Domain-agnostic training aims to enable
models to learn these common expressive patterns. To construct domain-agnostic
dialogue corpora, we employ a \textit{\textbf{de-domaining}} data processing
technique used to remove domain-specific features. By mitigating the effects of
domain-specific features, the model trained on the de-domained corpora can
effectively learn common expression patterns in different domains.
Subsequently, we adapt the learned domain-agnostic features to the target
domain through domain adaptation training. We conduct experiments on Chinese
dialogue datasets from five different domains and show that AMD$^2$G achieves
superior performance compared to both direct training on the target domain
corpus and collective training on all five domain corpora. Our work underscores
AMD$^2$G as a viable alternative solution for low-resource multi-domain
dialogue generation. Code and data associated with our work are available on
GitHub repository$^{\text 1}$. | cs.CL | A Unified Data Augmentation Framework for Low-Resource Multi-Domain Dialogue Generation |
2024-06-14T09:22:07Z | http://arxiv.org/pdf/2406.09864v1 | http://arxiv.org/abs/2406.09864v1 | Grigor Bezirganyan, Sana Sellami, Laure Berti-Équille, Sébastien Fournier | 2024-06-14T09:22:07Z | cs.LG, cs.AI, cs.CL, cs.CV | Multimodal Deep Learning enhances decision-making by integrating diverse
information sources, such as texts, images, audio, and videos. To develop
trustworthy multimodal approaches, it is essential to understand how
uncertainty impacts these models. We introduce LUMA, a unique benchmark
dataset, featuring audio, image, and textual data from 50 classes, for learning
from uncertain and multimodal data. It extends the well-known CIFAR 10/100
dataset with audio samples extracted from three audio corpora, and text data
generated using the Gemma-7B Large Language Model (LLM). The LUMA dataset
enables the controlled injection of varying types and degrees of uncertainty to
achieve and tailor specific experiments and benchmarking initiatives. LUMA is
also available as a Python package including the functions for generating
multiple variants of the dataset with controlling the diversity of the data,
the amount of noise for each modality, and adding out-of-distribution samples.
A baseline pre-trained model is also provided alongside three uncertainty
quantification methods: Monte-Carlo Dropout, Deep Ensemble, and Reliable
Conflictive Multi-View Learning. This comprehensive dataset and its tools are
intended to promote and support the development and benchmarking of trustworthy
and robust multimodal deep learning approaches. | cs.LG | LUMA: A Benchmark Dataset for Learning from Uncertain and Multimodal Data |
2024-06-14T09:10:24Z | http://arxiv.org/pdf/2406.09855v1 | http://arxiv.org/abs/2406.09855v1 | Aravind Krishnan, Badr M. Abdullah, Dietrich Klakow | 2024-06-14T09:10:24Z | cs.CL | While existing literature relies on performance differences to uncover gender
biases in ASR models, a deeper analysis is essential to understand how gender
is encoded and utilized during transcript generation. This work investigates
the encoding and utilization of gender in the latent representations of two
transformer-based ASR models, Wav2Vec2 and HuBERT. Using linear erasure, we
demonstrate the feasibility of removing gender information from each layer of
an ASR model and show that such an intervention has minimal impacts on the ASR
performance. Additionally, our analysis reveals a concentration of gender
information within the first and last frames in the final layers, explaining
the ease of erasing gender in these layers. Our findings suggest the prospect
of creating gender-neutral embeddings that can be integrated into ASR
frameworks without compromising their efficacy. | cs.CL | On the Encoding of Gender in Transformer-based ASR Representations |
2024-06-14T08:47:15Z | http://arxiv.org/pdf/2406.09839v1 | http://arxiv.org/abs/2406.09839v1 | Muhammad Yeza Baihaqi, Angel García Contreras, Seiya Kawano, Koichiro Yoshino | 2024-06-14T08:47:15Z | cs.CL, cs.HC | Rapport is known as a conversational aspect focusing on relationship
building, which influences outcomes in collaborative tasks. This study aims to
establish human-agent rapport through small talk by using a rapport-building
strategy. We implemented this strategy for the virtual agents based on dialogue
strategies by prompting a large language model (LLM). In particular, we
utilized two dialogue strategies-predefined sequence and free-form-to guide the
dialogue generation framework. We conducted analyses based on human
evaluations, examining correlations between total turn, utterance characters,
rapport score, and user experience variables: naturalness, satisfaction,
interest, engagement, and usability. We investigated correlations between
rapport score and naturalness, satisfaction, engagement, and conversation flow.
Our experimental results also indicated that using free-form to prompt the
rapport-building strategy performed the best in subjective scores. | cs.CL | Rapport-Driven Virtual Agent: Rapport Building Dialogue Strategy for Improving User Experience at First Meeting |
2024-06-14T08:40:58Z | http://arxiv.org/pdf/2406.09831v1 | http://arxiv.org/abs/2406.09831v1 | Youyang Qu | 2024-06-14T08:40:58Z | cs.LG, cs.AI, cs.CL, cs.NE | Federated learning (FL) offers a compelling framework for training large
language models (LLMs) while addressing data privacy and decentralization
challenges. This paper surveys recent advancements in the federated learning of
large language models, with a particular focus on machine unlearning, a crucial
aspect for complying with privacy regulations like the Right to be Forgotten.
Machine unlearning in the context of federated LLMs involves systematically and
securely removing individual data contributions from the learned model without
retraining from scratch. We explore various strategies that enable effective
unlearning, such as perturbation techniques, model decomposition, and
incremental learning, highlighting their implications for maintaining model
performance and data privacy. Furthermore, we examine case studies and
experimental results from recent literature to assess the effectiveness and
efficiency of these approaches in real-world scenarios. Our survey reveals a
growing interest in developing more robust and scalable federated unlearning
methods, suggesting a vital area for future research in the intersection of AI
ethics and distributed machine learning technologies. | cs.LG | Federated Learning driven Large Language Models for Swarm Intelligence: A Survey |
2024-06-14T08:32:45Z | http://arxiv.org/pdf/2406.09827v1 | http://arxiv.org/abs/2406.09827v1 | Heejun Lee, Geon Park, Youngwan Lee, Jina Kim, Wonyoung Jeong, Myeongjae Jeon, Sung Ju Hwang | 2024-06-14T08:32:45Z | cs.CL, cs.CV, cs.DC, cs.LG | In modern large language models (LLMs), increasing sequence lengths is a
crucial challenge for enhancing their comprehension and coherence in handling
complex tasks such as multi-modal question answering. However, handling long
context sequences with LLMs is prohibitively costly due to the conventional
attention mechanism's quadratic time and space complexity, and the context
window size is limited by the GPU memory. Although recent works have proposed
linear and sparse attention mechanisms to address this issue, their real-world
applicability is often limited by the need to re-train pre-trained models. In
response, we propose a novel approach, Hierarchically Pruned Attention (HiP),
which simultaneously reduces the training and inference time complexity from
$O(T^2)$ to $O(T \log T)$ and the space complexity from $O(T^2)$ to $O(T)$. To
this end, we devise a dynamic sparse attention mechanism that generates an
attention mask through a novel tree-search-like algorithm for a given query on
the fly. HiP is training-free as it only utilizes the pre-trained attention
scores to spot the positions of the top-$k$ most significant elements for each
query. Moreover, it ensures that no token is overlooked, unlike the sliding
window-based sub-quadratic attention methods, such as StreamingLLM. Extensive
experiments on diverse real-world benchmarks demonstrate that HiP significantly
reduces prompt (i.e., prefill) and decoding latency and memory usage while
maintaining high generation performance with little or no degradation. As HiP
allows pretrained LLMs to scale to millions of tokens on commodity GPUs with no
additional engineering due to its easy plug-and-play deployment, we believe
that our work will have a large practical impact, opening up the possibility to
many long-context LLM applications previously infeasible. | cs.CL | HiP Attention: Sparse Sub-Quadratic Attention with Hierarchical Attention Pruning |
2024-06-14T08:17:44Z | http://arxiv.org/pdf/2406.15465v1 | http://arxiv.org/abs/2406.15465v1 | Daniel Reichenpfader, Jonas Knupp, André Sander, Kerstin Denecke | 2024-06-14T08:17:44Z | cs.CL, cs.AI, J.3 | Annually and globally, over three billion radiography examinations and
computer tomography scans result in mostly unstructured radiology reports
containing free text. Despite the potential benefits of structured reporting,
its adoption is limited by factors such as established processes, resource
constraints and potential loss of information. However, structured information
would be necessary for various use cases, including automatic analysis,
clinical trial matching, and prediction of health outcomes. This study
introduces RadEx, an end-to-end framework comprising 15 software components and
ten artifacts to develop systems that perform automated information extraction
from radiology reports. It covers the complete process from annotating training
data to extracting information by offering a consistent generic information
model and setting boundaries for model development. Specifically, RadEx allows
clinicians to define relevant information for clinical domains (e.g.,
mammography) and to create report templates. The framework supports both
generative and encoder-only models and the decoupling of information extraction
from template filling enables independent model improvements. Developing
information extraction systems according to the RadEx framework facilitates
implementation and maintenance as components are easily exchangeable, while
standardized artifacts ensure interoperability between components. | cs.CL | RadEx: A Framework for Structured Information Extraction from Radiology Reports based on Large Language Models |
2024-06-14T08:13:34Z | http://arxiv.org/pdf/2406.09815v1 | http://arxiv.org/abs/2406.09815v1 | Zhenrui Yue, Huimin Zeng, Lanyu Shang, Yifan Liu, Yang Zhang, Dong Wang | 2024-06-14T08:13:34Z | cs.CL, cs.AI | The rapid propagation of misinformation poses substantial risks to public
interest. To combat misinformation, large language models (LLMs) are adapted to
automatically verify claim credibility. Nevertheless, existing methods heavily
rely on the embedded knowledge within LLMs and / or black-box APIs for evidence
collection, leading to subpar performance with smaller LLMs or upon unreliable
context. In this paper, we propose retrieval augmented fact verification
through the synthesis of contrasting arguments (RAFTS). Upon input claims,
RAFTS starts with evidence retrieval, where we design a retrieval pipeline to
collect and re-rank relevant documents from verifiable sources. Then, RAFTS
forms contrastive arguments (i.e., supporting or refuting) conditioned on the
retrieved evidence. In addition, RAFTS leverages an embedding model to identify
informative demonstrations, followed by in-context prompting to generate the
prediction and explanation. Our method effectively retrieves relevant documents
as evidence and evaluates arguments from varying perspectives, incorporating
nuanced information for fine-grained decision-making. Combined with informative
in-context examples as prior, RAFTS achieves significant improvements to
supervised and LLM baselines without complex prompts. We demonstrate the
effectiveness of our method through extensive experiments, where RAFTS can
outperform GPT-based methods with a significantly smaller 7B LLM. | cs.CL | Retrieval Augmented Fact Verification by Synthesizing Contrastive Arguments |
2024-06-14T07:47:09Z | http://arxiv.org/pdf/2406.15126v1 | http://arxiv.org/abs/2406.15126v1 | Lin Long, Rui Wang, Ruixuan Xiao, Junbo Zhao, Xiao Ding, Gang Chen, Haobo Wang | 2024-06-14T07:47:09Z | cs.CL | Within the evolving landscape of deep learning, the dilemma of data quantity
and quality has been a long-standing problem. The recent advent of Large
Language Models (LLMs) offers a data-centric solution to alleviate the
limitations of real-world data with synthetic data generation. However, current
investigations into this field lack a unified framework and mostly stay on the
surface. Therefore, this paper provides an organization of relevant studies
based on a generic workflow of synthetic data generation. By doing so, we
highlight the gaps within existing research and outline prospective avenues for
future study. This work aims to shepherd the academic and industrial
communities towards deeper, more methodical inquiries into the capabilities and
applications of LLMs-driven synthetic data generation. | cs.CL | On LLMs-Driven Synthetic Data Generation, Curation, and Evaluation: A Survey |
2024-06-14T07:40:07Z | http://arxiv.org/pdf/2406.09790v1 | http://arxiv.org/abs/2406.09790v1 | Bowen Zhang, Chunping Li | 2024-06-14T07:40:07Z | cs.CL | Semantic Textual Similarity (STS) constitutes a critical research direction
in computational linguistics and serves as a key indicator of the encoding
capabilities of embedding models. Driven by advances in pre-trained language
models and contrastive learning techniques, leading sentence representation
methods can already achieved average Spearman's correlation scores of
approximately 86 across seven STS benchmarks in SentEval. However, further
improvements have become increasingly marginal, with no existing method
attaining an average score higher than 87 on these tasks. This paper conducts
an in-depth analysis of this phenomenon and concludes that the upper limit for
Spearman's correlation scores using contrastive learning is 87.5. To transcend
this ceiling, we propose an innovative approach termed Pcc-tuning, which
employs Pearson's correlation coefficient as a loss function to refine model
performance beyond contrastive learning. Experimental results demonstrate that
Pcc-tuning markedly surpasses previous state-of-the-art strategies, raising the
Spearman's correlation score to above 90. | cs.CL | Pcc-tuning: Breaking the Contrastive Learning Ceiling in Semantic Textual Similarity |
2024-06-14T07:28:02Z | http://arxiv.org/pdf/2406.09779v1 | http://arxiv.org/abs/2406.09779v1 | Jingtao Cao, Zheng Zhang, Hongru Wang, Bin Liang, Hao Wang, Kam-Fai Wong | 2024-06-14T07:28:02Z | cs.AI, cs.CL, cs.CV | Memes, which rapidly disseminate personal opinions and positions across the
internet, also pose significant challenges in propagating social bias and
prejudice. This study presents a novel approach to detecting harmful memes,
particularly within the multicultural and multilingual context of Singapore.
Our methodology integrates image captioning, Optical Character Recognition
(OCR), and Large Language Model (LLM) analysis to comprehensively understand
and classify harmful memes. Utilizing the BLIP model for image captioning,
PP-OCR and TrOCR for text recognition across multiple languages, and the Qwen
LLM for nuanced language understanding, our system is capable of identifying
harmful content in memes created in English, Chinese, Malay, and Tamil. To
enhance the system's performance, we fine-tuned our approach by leveraging
additional data labeled using GPT-4V, aiming to distill the understanding
capability of GPT-4V for harmful memes to our system. Our framework achieves
top-1 at the public leaderboard of the Online Safety Prize Challenge hosted by
AI Singapore, with the AUROC as 0.7749 and accuracy as 0.7087, significantly
ahead of the other teams. Notably, our approach outperforms previous
benchmarks, with FLAVA achieving an AUROC of 0.5695 and VisualBERT an AUROC of
0.5561. | cs.AI | OSPC: Detecting Harmful Memes with Large Language Model as a Catalyst |
2024-06-14T07:06:24Z | http://arxiv.org/pdf/2406.09765v2 | http://arxiv.org/abs/2406.09765v2 | Liyang Wang, Yu Cheng, Ao Xiang, Jingyu Zhang, Haowei Yang | 2024-06-20T13:12:23Z | q-fin.RM, cs.CL | This paper explores the application of Natural Language Processing (NLP) in
financial risk detection. By constructing an NLP-based financial risk detection
model, this study aims to identify and predict potential risks in financial
documents and communications. First, the fundamental concepts of NLP and its
theoretical foundation, including text mining methods, NLP model design
principles, and machine learning algorithms, are introduced. Second, the
process of text data preprocessing and feature extraction is described.
Finally, the effectiveness and predictive performance of the model are
validated through empirical research. The results show that the NLP-based
financial risk detection model performs excellently in risk identification and
prediction, providing effective risk management tools for financial
institutions. This study offers valuable references for the field of financial
risk management, utilizing advanced NLP techniques to improve the accuracy and
efficiency of financial risk detection. | q-fin.RM | Application of Natural Language Processing in Financial Risk Detection |
2024-06-14T06:57:18Z | http://arxiv.org/pdf/2406.09760v1 | http://arxiv.org/abs/2406.09760v1 | Changyu Chen, Zichen Liu, Chao Du, Tianyu Pang, Qian Liu, Arunesh Sinha, Pradeep Varakantham, Min Lin | 2024-06-14T06:57:18Z | cs.CL, cs.LG | Human alignment in large language models (LLMs) is an active area of
research. A recent groundbreaking work, direct preference optimization (DPO),
has greatly simplified the process from past work in reinforcement learning
from human feedback (RLHF) by bypassing the reward learning stage in RLHF. DPO,
after training, provides an implicit reward model. In this work, we make a
novel observation that this implicit reward model can by itself be used in a
bootstrapping fashion to further align the LLM. Our approach is to use the
rewards from a current LLM model to construct a preference dataset, which is
then used in subsequent DPO rounds. We incorporate refinements that debias the
length of the responses and improve the quality of the preference dataset to
further improve our approach. Our approach, named self-alignment with DPO
ImpliCit rEwards (DICE), shows great improvements in alignment and achieves
superior performance than Gemini Pro on AlpacaEval 2, reaching 27.55%
length-controlled win rate against GPT-4 Turbo, but with only 8B parameters and
no external feedback. Our code is available at https://github.com/sail-sg/dice. | cs.CL | Bootstrapping Language Models with DPO Implicit Rewards |
2024-06-14T06:47:40Z | http://arxiv.org/pdf/2406.10311v1 | http://arxiv.org/abs/2406.10311v1 | Wenjing Zhang, Xuejiao Lei, Zhaoxiang Liu, Meijuan An, Bikun Yang, KaiKai Zhao, Kai Wang, Shiguo Lian | 2024-06-14T06:47:40Z | cs.CL, cs.AI | With the profound development of large language models(LLMs), their safety
concerns have garnered increasing attention. However, there is a scarcity of
Chinese safety benchmarks for LLMs, and the existing safety taxonomies are
inadequate, lacking comprehensive safety detection capabilities in authentic
Chinese scenarios. In this work, we introduce CHiSafetyBench, a dedicated
safety benchmark for evaluating LLMs' capabilities in identifying risky content
and refusing answering risky questions in Chinese contexts. CHiSafetyBench
incorporates a dataset that covers a hierarchical Chinese safety taxonomy
consisting of 5 risk areas and 31 categories. This dataset comprises two types
of tasks: multiple-choice questions and question-answering, evaluating LLMs
from the perspectives of risk content identification and the ability to refuse
answering risky questions respectively. Utilizing this benchmark, we validate
the feasibility of automatic evaluation as a substitute for human evaluation
and conduct comprehensive automatic safety assessments on mainstream Chinese
LLMs. Our experiments reveal the varying performance of different models across
various safety domains, indicating that all models possess considerable
potential for improvement in Chinese safety capabilities. Our dataset is
publicly available at
https://github.com/UnicomAI/DataSet/tree/main/TestData/Safety. | cs.CL | CHiSafetyBench: A Chinese Hierarchical Safety Benchmark for Large Language Models |
2024-06-14T06:22:47Z | http://arxiv.org/pdf/2406.10310v1 | http://arxiv.org/abs/2406.10310v1 | Zhuofeng Li, Zixing Gou, Xiangnan Zhang, Zhongyuan Liu, Sirui Li, Yuntong Hu, Chen Ling, Zheng Zhang, Liang Zhao | 2024-06-14T06:22:47Z | cs.CL, cs.AI | Text-Attributed Graphs (TAGs) augment graph structures with natural language
descriptions, facilitating detailed depictions of data and their
interconnections across various real-world settings. However, existing TAG
datasets predominantly feature textual information only at the nodes, with
edges typically represented by mere binary or categorical attributes. This lack
of rich textual edge annotations significantly limits the exploration of
contextual relationships between entities, hindering deeper insights into
graph-structured data. To address this gap, we introduce Textual-Edge Graphs
Datasets and Benchmark (TEG-DB), a comprehensive and diverse collection of
benchmark textual-edge datasets featuring rich textual descriptions on nodes
and edges. The TEG-DB datasets are large-scale and encompass a wide range of
domains, from citation networks to social networks. In addition, we conduct
extensive benchmark experiments on TEG-DB to assess the extent to which current
techniques, including pre-trained language models, graph neural networks, and
their combinations, can utilize textual node and edge information. Our goal is
to elicit advancements in textual-edge graph research, specifically in
developing methodologies that exploit rich textual node and edge descriptions
to enhance graph analysis and provide deeper insights into complex real-world
networks. The entire TEG-DB project is publicly accessible as an open-source
repository on Github, accessible at
https://github.com/Zhuofeng-Li/TEG-Benchmark. | cs.CL | TEG-DB: A Comprehensive Dataset and Benchmark of Textual-Edge Graphs |
2024-06-14T05:11:32Z | http://arxiv.org/pdf/2406.09719v1 | http://arxiv.org/abs/2406.09719v1 | Hancheol Park, Soyeong Jeong, Sukmin Cho, Jong C. Park | 2024-06-14T05:11:32Z | cs.CL, cs.AI | Recent language models have shown remarkable performance on natural language
understanding (NLU) tasks. However, they are often sub-optimal when faced with
ambiguous samples that can be interpreted in multiple ways, over-confidently
predicting a single label without consideration for its correctness. To address
this issue, we propose a novel self-knowledge distillation method that enables
models to learn label distributions more accurately by leveraging knowledge
distilled from their lower layers. This approach also includes a learning phase
that re-calibrates the unnecessarily strengthened confidence for training
samples judged as extremely ambiguous based on the distilled distribution
knowledge. We validate our method on diverse NLU benchmark datasets and the
experimental results demonstrate its effectiveness in producing better label
distributions. Particularly, through the process of re-calibrating the
confidence for highly ambiguous samples, the issue of over-confidence when
predictions for unseen samples do not match with their ground-truth labels has
been significantly alleviated. This has been shown to contribute to generating
better distributions than the existing state-of-the-art method. Moreover, our
method is more efficient in training the models compared to the existing
method, as it does not involve additional training processes to refine label
distributions. | cs.CL | Self-Knowledge Distillation for Learning Ambiguity |
2024-06-14T04:55:30Z | http://arxiv.org/pdf/2406.09717v2 | http://arxiv.org/abs/2406.09717v2 | Trinh Pham, Khoi M. Le, Luu Anh Tuan | 2024-06-17T04:05:39Z | cs.CL | In this paper, we introduce UniBridge (Cross-Lingual Transfer Learning with
Optimized Embeddings and Vocabulary), a comprehensive approach developed to
improve the effectiveness of Cross-Lingual Transfer Learning, particularly in
languages with limited resources. Our approach tackles two essential elements
of a language model: the initialization of embeddings and the optimal
vocabulary size. Specifically, we propose a novel embedding initialization
method that leverages both lexical and semantic alignment for a language. In
addition, we present a method for systematically searching for the optimal
vocabulary size, ensuring a balance between model complexity and linguistic
coverage. Our experiments across multilingual datasets show that our approach
greatly improves the F1-Score in several languages. UniBridge is a robust and
adaptable solution for cross-lingual systems in various languages, highlighting
the significance of initializing embeddings and choosing the right vocabulary
size in cross-lingual environments. | cs.CL | UniBridge: A Unified Approach to Cross-Lingual Transfer Learning for Low-Resource Languages |
2024-06-14T04:52:15Z | http://arxiv.org/pdf/2406.10307v1 | http://arxiv.org/abs/2406.10307v1 | Shiguo Lian, Kaikai Zhao, Xinhui Liu, Xuejiao Lei, Bikun Yang, Wenjing Zhang, Kai Wang, Zhaoxiang Liu | 2024-06-14T04:52:15Z | cs.CL, cs.AI | General large language models enhanced with supervised fine-tuning and
reinforcement learning from human feedback are increasingly popular in academia
and industry as they generalize foundation models to various practical tasks in
a prompt manner. To assist users in selecting the best model in practical
application scenarios, i.e., choosing the model that meets the application
requirements while minimizing cost, we introduce A-Eval, an application-driven
LLMs evaluation benchmark for general large language models. First, we
categorize evaluation tasks into five main categories and 27 sub-categories
from a practical application perspective. Next, we construct a dataset
comprising 678 question-and-answer pairs through a process of collecting,
annotating, and reviewing. Then, we design an objective and effective
evaluation method and evaluate a series of LLMs of different scales on A-Eval.
Finally, we reveal interesting laws regarding model scale and task difficulty
level and propose a feasible method for selecting the best model. Through
A-Eval, we provide clear empirical and engineer guidance for selecting the best
model, reducing barriers to selecting and using LLMs and promoting their
application and development. Our benchmark is publicly available at
https://github.com/UnicomAI/DataSet/tree/main/TestData/GeneralAbility. | cs.CL | What is the best model? Application-driven Evaluation for Large Language Models |
2024-06-14T04:03:24Z | http://arxiv.org/pdf/2406.09702v1 | http://arxiv.org/abs/2406.09702v1 | Ryohei Kamei, Daiki Shiono, Reina Akama, Jun Suzuki | 2024-06-14T04:03:24Z | cs.CL | With the remarkable development of large language models (LLMs), ensuring the
factuality of output has become a challenge. However, having all the contents
of the response with given knowledge or facts is not necessarily a good thing
in dialogues. This study aimed to achieve both attractiveness and factuality in
a dialogue response for which a task was set to predict sentences that do not
require factual correctness judgment such as agreeing, or personal
opinions/feelings. We created a dataset, dialogue dataset annotated with
fact-check-needed label (DDFC), for this task via crowdsourcing, and
classification tasks were performed on several models using this dataset. The
model with the highest classification accuracy could yield about 88% accurate
classification results. | cs.CL | Detecting Response Generation Not Requiring Factual Judgment |
2024-06-14T03:34:02Z | http://arxiv.org/pdf/2406.11890v1 | http://arxiv.org/abs/2406.11890v1 | Hui Liu, Wenya Wang, Hao Sun, Chris Xing Tian, Chenqi Kong, Xin Dong, Haoliang Li | 2024-06-14T03:34:02Z | cs.LG, cs.AI, cs.CL | Large Language Models (LLMs) have demonstrated impressive in-context learning
(ICL) capabilities from few-shot demonstration exemplars. While recent
learning-based demonstration selection methods have proven beneficial to ICL by
choosing more useful exemplars, their underlying mechanisms are opaque,
hindering efforts to address limitations such as high training costs and poor
generalization across tasks. These methods generally assume the selection
process captures similarities between the exemplar and the target instance,
however, it remains unknown what kinds of similarities are captured and vital
to performing ICL. To dive into this question, we analyze the working
mechanisms of the learning-based demonstration selection methods and
empirically identify two important factors related to similarity measurement:
1) The ability to integrate different levels of task-agnostic text similarities
between the input of exemplars and test cases enhances generalization power
across different tasks. 2) Incorporating task-specific labels when measuring
the similarities significantly improves the performance on each specific task.
We validate these two findings through extensive quantitative and qualitative
analyses across ten datasets and various LLMs. Based on our findings, we
introduce two effective yet simplified exemplar selection methods catering to
task-agnostic and task-specific demands, eliminating the costly LLM inference
overhead. | cs.LG | Unraveling the Mechanics of Learning-Based Demonstration Selection for In-Context Learning |
2024-06-14T03:18:28Z | http://arxiv.org/pdf/2406.09688v1 | http://arxiv.org/abs/2406.09688v1 | Zijian Feng, Hanzhang Zhou, Zixiao Zhu, Kezhi Mao | 2024-06-14T03:18:28Z | cs.CL | Controllable text generation (CTG) seeks to craft texts adhering to specific
attributes, traditionally employing learning-based techniques such as training,
fine-tuning, or prefix-tuning with attribute-specific datasets. These
approaches, while effective, demand extensive computational and data resources.
In contrast, some proposed learning-free alternatives circumvent learning but
often yield inferior results, exemplifying the fundamental machine learning
trade-off between computational expense and model efficacy. To overcome these
limitations, we propose FreeCtrl, a learning-free approach that dynamically
adjusts the weights of selected feedforward neural network (FFN) vectors to
steer the outputs of large language models (LLMs). FreeCtrl hinges on the
principle that the weights of different FFN vectors influence the likelihood of
different tokens appearing in the output. By identifying and adaptively
adjusting the weights of attribute-related FFN vectors, FreeCtrl can control
the output likelihood of attribute keywords in the generated content. Extensive
experiments on single- and multi-attribute control reveal that the
learning-free FreeCtrl outperforms other learning-free and learning-based
methods, successfully resolving the dilemma between learning costs and model
performance. | cs.CL | FreeCtrl: Constructing Control Centers with Feedforward Layers for Learning-Free Controllable Text Generation |
2024-06-14T03:06:55Z | http://arxiv.org/pdf/2406.10304v1 | http://arxiv.org/abs/2406.10304v1 | Ming Gao, Hang Chen, Jun Du, Xin Xu, Hongxiao Guo, Hui Bu, Jianxing Yang, Ming Li, Chin-Hui Lee | 2024-06-14T03:06:55Z | cs.CL | Smart home technology has gained widespread adoption, facilitating effortless
control of devices through voice commands. However, individuals with
dysarthria, a motor speech disorder, face challenges due to the variability of
their speech. This paper addresses the wake-up word spotting (WWS) task for
dysarthric individuals, aiming to integrate them into real-world applications.
To support this, we release the open-source Mandarin Dysarthria Speech Corpus
(MDSC), a dataset designed for dysarthric individuals in home environments.
MDSC encompasses information on age, gender, disease types, and intelligibility
evaluations. Furthermore, we perform comprehensive experimental analysis on
MDSC, highlighting the challenges encountered. We also develop a customized
dysarthria WWS system that showcases robustness in handling intelligibility and
achieving exceptional performance. MDSC will be released on
https://www.aishelltech.com/AISHELL_6B. | cs.CL | Enhancing Voice Wake-Up for Dysarthria: Mandarin Dysarthria Speech Corpus Release and Customized System Design |
2024-06-14T02:58:19Z | http://arxiv.org/pdf/2406.09676v1 | http://arxiv.org/abs/2406.09676v1 | Roger Hsiao, Liuhui Deng, Erik McDermott, Ruchir Travadi, Xiaodan Zhuang | 2024-06-14T02:58:19Z | eess.AS, cs.CL | We propose a novel approach to optimizing a byte-level representation for
end-to-end automatic speech recognition (ASR). Byte-level representation is
often used by large scale multilingual ASR systems when the character set of
the supported languages is large. The compactness and universality of
byte-level representation allow the ASR models to use smaller output
vocabularies and therefore, provide more flexibility. UTF-8 is a commonly used
byte-level representation for multilingual ASR, but it is not designed to
optimize machine learning tasks directly. By using auto-encoder and vector
quantization, we show that we can optimize a byte-level representation for ASR
and achieve better accuracy. Our proposed framework can incorporate information
from different modalities, and provides an error correction mechanism. In an
English/Mandarin dictation task, we show that a bilingual ASR model built with
this approach can outperform UTF-8 representation by 5% relative in error rate. | eess.AS | Optimizing Byte-level Representation for End-to-end ASR |
2024-06-14T02:42:30Z | http://arxiv.org/pdf/2406.09671v1 | http://arxiv.org/abs/2406.09671v1 | Nabor C. Mendonça | 2024-06-14T02:42:30Z | cs.AI, cs.CL | The recent integration of visual capabilities into Large Language Models
(LLMs) has the potential to play a pivotal role in science and technology
education, where visual elements such as diagrams, charts, and tables are
commonly used to improve the learning experience. This study investigates the
performance of ChatGPT-4 Vision, OpenAI's most advanced visual model at the
time the study was conducted, on the Bachelor in Computer Science section of
Brazil's 2021 National Undergraduate Exam (ENADE). By presenting the model with
the exam's open and multiple-choice questions in their original image format
and allowing for reassessment in response to differing answer keys, we were
able to evaluate the model's reasoning and self-reflecting capabilities in a
large-scale academic assessment involving textual and visual content. ChatGPT-4
Vision significantly outperformed the average exam participant, positioning
itself within the top 10 best score percentile. While it excelled in questions
that incorporated visual elements, it also encountered challenges with question
interpretation, logical reasoning, and visual acuity. The involvement of an
independent expert panel to review cases of disagreement between the model and
the answer key revealed some poorly constructed questions containing vague or
ambiguous statements, calling attention to the critical need for improved
question design in future exams. Our findings suggest that while ChatGPT-4
Vision shows promise in multimodal academic evaluations, human oversight
remains crucial for verifying the model's accuracy and ensuring the fairness of
high-stakes educational exams. The paper's research materials are publicly
available at https://github.com/nabormendonca/gpt-4v-enade-cs-2021. | cs.AI | Evaluating ChatGPT-4 Vision on Brazil's National Undergraduate Computer Science Exam |
2024-06-14T02:42:20Z | http://arxiv.org/pdf/2406.10303v1 | http://arxiv.org/abs/2406.10303v1 | Jinqiang Wang, Huansheng Ning, Yi Peng, Qikai Wei, Daniel Tesfai, Wenwei Mao, Tao Zhu, Runhe Huang | 2024-06-14T02:42:20Z | cs.CL, cs.AI | Large Language Models (LLMs) have demonstrated surprising performance across
various natural language processing tasks. Recently, medical LLMs enhanced with
domain-specific knowledge have exhibited excellent capabilities in medical
consultation and diagnosis. These models can smoothly simulate doctor-patient
dialogues and provide professional medical advice. Most medical LLMs are
developed through continued training of open-source general LLMs, which require
significantly fewer computational resources than training LLMs from scratch.
Additionally, this approach offers better protection of patient privacy
compared to API-based solutions. This survey systematically explores how to
train medical LLMs based on general LLMs. It covers: (a) how to acquire
training corpus and construct customized medical training sets, (b) how to
choose a appropriate training paradigm, (c) how to choose a suitable evaluation
benchmark, and (d) existing challenges and promising future research directions
are discussed. This survey can provide guidance for the development of LLMs
focused on various medical applications, such as medical education, diagnostic
planning, and clinical assistants. | cs.CL | A Survey on Large Language Models from General Purpose to Medical Applications: Datasets, Methodologies, and Evaluations |
2024-06-14T02:21:53Z | http://arxiv.org/pdf/2406.09662v1 | http://arxiv.org/abs/2406.09662v1 | Freda Shi | 2024-06-14T02:21:53Z | cs.CL, cs.AI, cs.CV | Language is highly structured, with syntactic and semantic structures, to
some extent, agreed upon by speakers of the same language. With implicit or
explicit awareness of such structures, humans can learn and use language
efficiently and generalize to sentences that contain unseen words. Motivated by
human language learning, in this dissertation, we consider a family of machine
learning tasks that aim to learn language structures through grounding. We seek
distant supervision from other data sources (i.e., grounds), including but not
limited to other modalities (e.g., vision), execution results of programs, and
other languages.
We demonstrate the potential of this task formulation and advocate for its
adoption through three schemes. In Part I, we consider learning syntactic
parses through visual grounding. We propose the task of visually grounded
grammar induction, present the first models to induce syntactic structures from
visually grounded text and speech, and find that the visual grounding signals
can help improve the parsing quality over language-only models. As a side
contribution, we propose a novel evaluation metric that enables the evaluation
of speech parsing without text or automatic speech recognition systems
involved. In Part II, we propose two execution-aware methods to map sentences
into corresponding semantic structures (i.e., programs), significantly
improving compositional generalization and few-shot program synthesis. In Part
III, we propose methods that learn language structures from annotations in
other languages. Specifically, we propose a method that sets a new state of the
art on cross-lingual word alignment. We then leverage the learned word
alignments to improve the performance of zero-shot cross-lingual dependency
parsing, by proposing a novel substructure-based projection method that
preserves structural knowledge learned from the source language. | cs.CL | Learning Language Structures through Grounding |
2024-06-13T22:55:22Z | http://arxiv.org/pdf/2406.09618v1 | http://arxiv.org/abs/2406.09618v1 | Jari Kolehmainen, Aditya Gourav, Prashanth Gurunath Shivakumar, Yile Gu, Ankur Gandhe, Ariya Rastrow, Grant Strimel, Ivan Bulyko | 2024-06-13T22:55:22Z | cs.CL, cs.AI, cs.IR, cs.SD, eess.AS | Retrieval is a widely adopted approach for improving language models
leveraging external information. As the field moves towards multi-modal large
language models, it is important to extend the pure text based methods to
incorporate other modalities in retrieval as well for applications across the
wide spectrum of machine learning tasks and data types. In this work, we
propose multi-modal retrieval with two approaches: kNN-LM and cross-attention
techniques. We demonstrate the effectiveness of our retrieval approaches
empirically by applying them to automatic speech recognition tasks with access
to external information. Under this setting, we show that speech-based
multi-modal retrieval outperforms text based retrieval, and yields up to 50 %
improvement in word error rate over the multi-modal language model baseline.
Furthermore, we achieve state-of-the-art recognition results on the
Spoken-Squad question answering dataset. | cs.CL | Multi-Modal Retrieval For Large Language Model Based Speech Recognition |
2024-06-13T22:52:07Z | http://arxiv.org/pdf/2406.09617v1 | http://arxiv.org/abs/2406.09617v1 | Shruti Palaskar, Oggi Rudovic, Sameer Dharur, Florian Pesce, Gautam Krishna, Aswin Sivaraman, Jack Berkowitz, Ahmed Hussen Abdelaziz, Saurabh Adya, Ahmed Tewfik | 2024-06-13T22:52:07Z | cs.CL, cs.HC, eess.AS | Although Large Language Models (LLMs) have shown promise for human-like
conversations, they are primarily pre-trained on text data. Incorporating audio
or video improves performance, but collecting large-scale multimodal data and
pre-training multimodal LLMs is challenging. To this end, we propose a Fusion
Low Rank Adaptation (FLoRA) technique that efficiently adapts a pre-trained
unimodal LLM to consume new, previously unseen modalities via low rank
adaptation. For device-directed speech detection, using FLoRA, the multimodal
LLM achieves 22% relative reduction in equal error rate (EER) over the
text-only approach and attains performance parity with its full fine-tuning
(FFT) counterpart while needing to tune only a fraction of its parameters.
Furthermore, with the newly introduced adapter dropout, FLoRA is robust to
missing data, improving over FFT by 20% lower EER and 56% lower false accept
rate. The proposed approach scales well for model sizes from 16M to 3B
parameters. | cs.CL | Multimodal Large Language Models with Fusion Low Rank Adaptation for Device Directed Speech Detection |
2024-06-13T21:32:56Z | http://arxiv.org/pdf/2406.10300v1 | http://arxiv.org/abs/2406.10300v1 | Irene Weber | 2024-06-13T21:32:56Z | cs.SE, cs.CL, cs.LG, A.1; I.2.7; D.2.11 | Large Language Models (LLMs) have become widely adopted recently. Research
explores their use both as autonomous agents and as tools for software
engineering. LLM-integrated applications, on the other hand, are software
systems that leverage an LLM to perform tasks that would otherwise be
impossible or require significant coding effort. While LLM-integrated
application engineering is emerging as new discipline, its terminology,
concepts and methods need to be established. This study provides a taxonomy for
LLM-integrated applications, offering a framework for analyzing and describing
these systems. It also demonstrates various ways to utilize LLMs in
applications, as well as options for implementing such integrations.
Following established methods, we analyze a sample of recent LLM-integrated
applications to identify relevant dimensions. We evaluate the taxonomy by
applying it to additional cases. This review shows that applications integrate
LLMs in numerous ways for various purposes. Frequently, they comprise multiple
LLM integrations, which we term ``LLM components''. To gain a clear
understanding of an application's architecture, we examine each LLM component
separately. We identify thirteen dimensions along which to characterize an LLM
component, including the LLM skills leveraged, the format of the output, and
more. LLM-integrated applications are described as combinations of their LLM
components. We suggest a concise representation using feature vectors for
visualization.
The taxonomy is effective for describing LLM-integrated applications. It can
contribute to theory building in the nascent field of LLM-integrated
application engineering and aid in developing such systems. Researchers and
practitioners explore numerous creative ways to leverage LLMs in applications.
Though challenges persist, integrating LLMs may revolutionize the way software
systems are built. | cs.SE | Large Language Models as Software Components: A Taxonomy for LLM-Integrated Applications |
2024-06-13T20:23:59Z | http://arxiv.org/pdf/2406.09573v1 | http://arxiv.org/abs/2406.09573v1 | Saba Yousefian Jazi, Amir Mirzaeinia, Sina Yousefian Jazi | 2024-06-13T20:23:59Z | cs.CL, cs.AI | In this effort we fine tuned different models based on BERT to detect the
gender polarity of twitter accounts. We specially focused on analyzing the
effect of using emojis and emoticons in performance of our model in classifying
task. We were able to demonstrate that the use of these none word inputs
alongside the mention of other accounts in a short text format like tweet has
an impact in detecting the account holder's gender. | cs.CL | Analyzing Gender Polarity in Short Social Media Texts with BERT: The Role of Emojis and Emoticons |
2024-06-13T20:20:29Z | http://arxiv.org/pdf/2406.09569v1 | http://arxiv.org/abs/2406.09569v1 | Frank Seide, Morrie Doulaty, Yangyang Shi, Yashesh Gaur, Junteng Jia, Chunyang Wu | 2024-06-13T20:20:29Z | cs.CL, cs.AI, cs.SD, eess.AS | We introduce Speech ReaLLM, a new ASR architecture that marries
"decoder-only" ASR with the RNN-T to make multimodal LLM architectures capable
of real-time streaming. This is the first "decoder-only" ASR architecture
designed to handle continuous audio without explicit end-pointing. Speech
ReaLLM is a special case of the more general ReaLLM ("real-time LLM") approach,
also introduced here for the first time. The idea is inspired by RNN-T: Instead
of generating a response only at the end of a user prompt, generate after every
input token received in real time (it is often empty). On Librispeech "test",
an 80M Speech ReaLLM achieves WERs of 3.0% and 7.4% in real time (without an
external LM or auxiliary loss). This is only slightly above a 3x larger
Attention-Encoder-Decoder baseline. We also show that this way, an LLM
architecture can learn to represent and reproduce the flow of time; and that a
pre-trained 7B LLM can be fine-tuned to do reasonably well on this task. | cs.CL | Speech ReaLLM -- Real-time Streaming Speech Recognition with Multimodal LLMs by Teaching the Flow of Time |
2024-06-13T19:55:20Z | http://arxiv.org/pdf/2406.09559v1 | http://arxiv.org/abs/2406.09559v1 | Sankalp KJ, Vinija Jain, Sreyoshi Bhaduri, Tamoghna Roy, Aman Chadha | 2024-06-13T19:55:20Z | cs.CL, cs.AI, cs.LG | This review paper provides a comprehensive overview of large language model
(LLM) research directions within Indic languages. Indic languages are those
spoken in the Indian subcontinent, including India, Pakistan, Bangladesh, Sri
Lanka, Nepal, and Bhutan, among others. These languages have a rich cultural
and linguistic heritage and are spoken by over 1.5 billion people worldwide.
With the tremendous market potential and growing demand for natural language
processing (NLP) based applications in diverse languages, generative
applications for Indic languages pose unique challenges and opportunities for
research. Our paper deep dives into the recent advancements in Indic generative
modeling, contributing with a taxonomy of research directions, tabulating 84
recent publications. Research directions surveyed in this paper include LLM
development, fine-tuning existing LLMs, development of corpora, benchmarking
and evaluation, as well as publications around specific techniques, tools, and
applications. We found that researchers across the publications emphasize the
challenges associated with limited data availability, lack of standardization,
and the peculiar linguistic complexities of Indic languages. This work aims to
serve as a valuable resource for researchers and practitioners working in the
field of NLP, particularly those focused on Indic languages, and contributes to
the development of more accurate and efficient LLM applications for these
languages. | cs.CL | Decoding the Diversity: A Review of the Indic AI Research Landscape |
2024-06-13T19:43:38Z | http://arxiv.org/pdf/2406.09556v2 | http://arxiv.org/abs/2406.09556v2 | Márton Kardos, Jan Kostkan, Arnault-Quentin Vermillet, Kristoffer Nielbo, Kenneth Enevoldsen, Roberta Rocca | 2024-06-18T14:12:18Z | cs.LG, cs.CL, stat.ML, I.2.7 | Topic models are useful tools for discovering latent semantic structures in
large textual corpora. Topic modeling historically relied on bag-of-words
representations of language. This approach makes models sensitive to the
presence of stop words and noise, and does not utilize potentially useful
contextual information. Recent efforts have been oriented at incorporating
contextual neural representations in topic modeling and have been shown to
outperform classical topic models. These approaches are, however, typically
slow, volatile and still require preprocessing for optimal results. We present
Semantic Signal Separation ($S^3$), a theory-driven topic modeling approach in
neural embedding spaces. $S^3$ conceptualizes topics as independent axes of
semantic space, and uncovers these with blind-source separation. Our approach
provides the most diverse, highly coherent topics, requires no preprocessing,
and is demonstrated to be the fastest contextually sensitive topic model to
date. We offer an implementation of $S^3$, among other approaches, in the
Turftopic Python package. | cs.LG | $S^3$ -- Semantic Signal Separation |
2024-06-13T19:30:32Z | http://arxiv.org/pdf/2406.09549v1 | http://arxiv.org/abs/2406.09549v1 | Nudrat Habib | 2024-06-13T19:30:32Z | cs.CL, cs.LG | Parsing is the process of breaking a sentence into its grammatical components
and identifying the syntactic structure of the sentence. The syntactically
correct sentence structure is achieved by assigning grammatical labels to its
constituents using lexicon and syntactic rules. In linguistics, parser is
extremely useful due to the number of different applications like name entity
recognition, QA systems and information extraction, etc. The two most common
techniques used for parsing are phrase structure and dependency Structure.
Because Urdu is a low-resource language, there has been little progress in
building an Urdu parser. A comparison of several parsers revealed that the
dependency parsing approach is better suited for order-free languages such as
Urdu. We have made significant progress in parsing Urdu, a South Asian language
with a complex morphology. For Urdu dependency parsing, a basic feature model
consisting of word location, word head, and dependency relation is employed as
a starting point, followed by more complex feature models. The dependency
tagset is designed after careful consideration of the complex morphological
structure of the Urdu language, word order variation, and lexical ambiguity and
it contains 22 tags. Our dataset comprises of sentences from news articles, and
we tried to include sentences of different complexity (which is quite
challenging), to get reliable results. All experiments are performed using
MaltParser, exploring all 9 algorithms and classifiers. We have achieved a 70
percent overall best-labeled accuracy (LA), as well as an 84 percent overall
best-unlabeled attachment score (UAS) using the Nivreeager algorithm. The
comparison of output data with treebank test data that has been manually parsed
is then used to carry out error assessment and to identify the errors produced
by the parser. | cs.CL | Exploring Syntactic Patterns in Urdu: A Deep Dive into Dependency Analysis |
2024-06-13T18:16:27Z | http://arxiv.org/pdf/2406.09520v1 | http://arxiv.org/abs/2406.09520v1 | Bayode Ogunleye, Kudirat Ibilola Zakariyyah, Oluwaseun Ajao, Olakunle Olayinka, Hemlata Sharma | 2024-06-13T18:16:27Z | cs.IR, cs.AI, cs.CL, cs.LG, H.3.3 | The use of generative artificial intelligence (GenAI) in academia is a
subjective and hotly debated topic. Currently, there are no agreed guidelines
towards the usage of GenAI systems in higher education (HE) and, thus, it is
still unclear how to make effective use of the technology for teaching and
learning practice. This paper provides an overview of the current state of
research on GenAI for teaching and learning in HE. To this end, this study
conducted a systematic review of relevant studies indexed by Scopus, using the
preferred reporting items for systematic reviews and meta-analyses (PRISMA)
guidelines. The search criteria revealed a total of 625 research papers, of
which 355 met the final inclusion criteria. The findings from the review showed
the current state and the future trends in documents, citations, document
sources/authors, keywords, and co-authorship. The research gaps identified
suggest that while some authors have looked at understanding the detection of
AI-generated text, it may be beneficial to understand how GenAI can be
incorporated into supporting the educational curriculum for assessments,
teaching, and learning delivery. Furthermore, there is a need for additional
interdisciplinary, multidimensional studies in HE through collaboration. This
will strengthen the awareness and understanding of students, tutors, and other
stakeholders, which will be instrumental in formulating guidelines, frameworks,
and policies for GenAI usage. | cs.IR | A Systematic Review of Generative AI for Teaching and Learning Practice |
2024-06-13T18:12:01Z | http://arxiv.org/pdf/2406.09519v1 | http://arxiv.org/abs/2406.09519v1 | Jack Merullo, Carsten Eickhoff, Ellie Pavlick | 2024-06-13T18:12:01Z | cs.CL, cs.AI | Although it is known that transformer language models (LMs) pass features
from early layers to later layers, it is not well understood how this
information is represented and routed by the model. By analyzing particular
mechanism LMs use to accomplish this, we find that it is also used to recall
items from a list, and show that this mechanism can explain an otherwise
arbitrary-seeming sensitivity of the model to the order of items in the prompt.
Specifically, we find that models write into low-rank subspaces of the residual
stream to represent features which are then read out by specific later layers,
forming low-rank communication channels between layers. By decomposing
attention head weight matrices with the Singular Value Decomposition (SVD), we
find that previously described interactions between heads separated by one or
more layers can be predicted via analysis of their weight matrices. We show
that it is possible to manipulate the internal model representations as well as
edit model weights based on the mechanism we discover in order to significantly
improve performance on our synthetic Laundry List task, which requires recall
from a list, often improving task accuracy by over 20%. Our analysis reveals a
surprisingly intricate interpretable structure learned from language model
pretraining, and helps us understand why sophisticated LMs sometimes fail in
simple domains, facilitating future analysis of more complex behaviors. | cs.CL | Talking Heads: Understanding Inter-layer Communication in Transformer Language Models |
2024-06-13T17:59:52Z | http://arxiv.org/pdf/2406.09411v2 | http://arxiv.org/abs/2406.09411v2 | Fei Wang, Xingyu Fu, James Y. Huang, Zekun Li, Qin Liu, Xiaogeng Liu, Mingyu Derek Ma, Nan Xu, Wenxuan Zhou, Kai Zhang, Tianyi Lorena Yan, Wenjie Jacky Mo, Hsiang-Hui Liu, Pan Lu, Chunyuan Li, Chaowei Xiao, Kai-Wei Chang, Dan Roth, Sheng Zhang, Hoifung Poon, Muhao Chen | 2024-07-02T01:56:14Z | cs.CV, cs.AI, cs.CL | We introduce MuirBench, a comprehensive benchmark that focuses on robust
multi-image understanding capabilities of multimodal LLMs. MuirBench consists
of 12 diverse multi-image tasks (e.g., scene understanding, ordering) that
involve 10 categories of multi-image relations (e.g., multiview, temporal
relations). Comprising 11,264 images and 2,600 multiple-choice questions,
MuirBench is created in a pairwise manner, where each standard instance is
paired with an unanswerable variant that has minimal semantic differences, in
order for a reliable assessment. Evaluated upon 20 recent multi-modal LLMs, our
results reveal that even the best-performing models like GPT-4o and Gemini Pro
find it challenging to solve MuirBench, achieving 68.0% and 49.3% in accuracy.
Open-source multimodal LLMs trained on single images can hardly generalize to
multi-image questions, hovering below 33.3% in accuracy. These results
highlight the importance of MuirBench in encouraging the community to develop
multimodal LLMs that can look beyond a single image, suggesting potential
pathways for future improvements. | cs.CV | MuirBench: A Comprehensive Benchmark for Robust Multi-image Understanding |
2024-06-13T17:59:31Z | http://arxiv.org/pdf/2406.09403v1 | http://arxiv.org/abs/2406.09403v1 | Yushi Hu, Weijia Shi, Xingyu Fu, Dan Roth, Mari Ostendorf, Luke Zettlemoyer, Noah A Smith, Ranjay Krishna | 2024-06-13T17:59:31Z | cs.CV, cs.CL | Humans draw to facilitate reasoning: we draw auxiliary lines when solving
geometry problems; we mark and circle when reasoning on maps; we use sketches
to amplify our ideas and relieve our limited-capacity working memory. However,
such actions are missing in current multimodal language models (LMs). Current
chain-of-thought and tool-use paradigms only use text as intermediate reasoning
steps. In this work, we introduce Sketchpad, a framework that gives multimodal
LMs a visual sketchpad and tools to draw on the sketchpad. The LM conducts
planning and reasoning according to the visual artifacts it has drawn.
Different from prior work, which uses text-to-image models to enable LMs to
draw, Sketchpad enables LMs to draw with lines, boxes, marks, etc., which is
closer to human sketching and better facilitates reasoning. Sketchpad can also
use specialist vision models during the sketching process (e.g., draw bounding
boxes with object detection models, draw masks with segmentation models), to
further enhance visual perception and reasoning. We experiment with a wide
range of math tasks (including geometry, functions, graphs, and chess) and
complex visual reasoning tasks. Sketchpad substantially improves performance on
all tasks over strong base models with no sketching, yielding an average gain
of 12.7% on math tasks, and 8.6% on vision tasks. GPT-4o with Sketchpad sets a
new state of the art on all tasks, including V*Bench (80.3%), BLINK spatial
reasoning (83.9%), and visual correspondence (80.8%). All codes and data are in
https://visualsketchpad.github.io/. | cs.CV | Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models |
2024-06-13T17:59:09Z | http://arxiv.org/pdf/2406.09393v1 | http://arxiv.org/abs/2406.09393v1 | Jianing Yang, Harshine Visvanathan, Yilin Wang, Xinyi Hu, Matthew Gormley | 2024-06-13T17:59:09Z | cs.CL, cs.AI, cs.LG | Many tasks within NLP can be framed as sequential decision problems, ranging
from sequence tagging to text generation. However, for many tasks, the standard
training methods, including maximum likelihood (teacher forcing) and scheduled
sampling, suffer from exposure bias and a mismatch between metrics employed
during training and inference. DAgger provides a solution to mitigate these
problems, yet it requires a metric-specific dynamic oracle algorithm, which
does not exist for many common metrics like span-based F1, ROUGE, and BLEU. In
this paper, we develop these novel dynamic oracles and show they maintain
DAgger's no-regret guarantee for decomposable metrics like span-based F1. We
evaluate the algorithm's performance on named entity recognition (NER), text
summarization, and machine translation (MT). While DAgger with dynamic oracle
yields less favorable results in our MT experiments, it outperforms the
baseline techniques in NER and text summarization. | cs.CL | Improving Autoregressive Training with Dynamic Oracles |
2024-06-13T17:28:13Z | http://arxiv.org/pdf/2406.09345v1 | http://arxiv.org/abs/2406.09345v1 | Suwon Shon, Kwangyoun Kim, Yi-Te Hsu, Prashant Sridhar, Shinji Watanabe, Karen Livescu | 2024-06-13T17:28:13Z | cs.CL, cs.SD, eess.AS | The integration of pre-trained text-based large language models (LLM) with
speech input has enabled instruction-following capabilities for diverse speech
tasks. This integration requires the use of a speech encoder, a speech adapter,
and an LLM, trained on diverse tasks. We propose the use of discrete speech
units (DSU), rather than continuous-valued speech encoder outputs, that are
converted to the LLM token embedding space using the speech adapter. We
generate DSU using a self-supervised speech encoder followed by k-means
clustering. The proposed model shows robust performance on speech inputs from
seen/unseen domains and instruction-following capability in spoken question
answering. We also explore various types of DSU extracted from different layers
of the self-supervised speech encoder, as well as Mel frequency Cepstral
Coefficients (MFCC). Our findings suggest that the ASR task and datasets are
not crucial in instruction-tuning for spoken question answering tasks. | cs.CL | DiscreteSLU: A Large Language Model with Self-Supervised Discrete Speech Units for Spoken Language Understanding |
2024-06-13T17:15:33Z | http://arxiv.org/pdf/2406.09334v2 | http://arxiv.org/abs/2406.09334v2 | David Anugraha, Genta Indra Winata, Chenyue Li, Patrick Amadeus Irawan, En-Shiun Annie Lee | 2024-06-14T14:52:05Z | cs.CL | Performance prediction is a method to estimate the performance of Language
Models (LMs) on various Natural Language Processing (NLP) tasks, mitigating
computational costs associated with model capacity and data for fine-tuning.
Our paper introduces ProxyLM, a scalable framework for predicting LM
performance using proxy models in multilingual tasks. These proxy models act as
surrogates, approximating the performance of the LM of interest. By leveraging
proxy models, ProxyLM significantly reduces computational overhead on task
evaluations, achieving up to a 37.08x speedup compared to traditional methods,
even with our smallest proxy models. Additionally, our methodology showcases
adaptability to previously unseen languages in pre-trained LMs, outperforming
the state-of-the-art performance by 1.89x as measured by root-mean-square error
(RMSE). This framework streamlines model selection, enabling efficient
deployment and iterative LM enhancements without extensive computational
resources. | cs.CL | ProxyLM: Predicting Language Model Performance on Multilingual Tasks via Proxy Models |
2024-06-13T17:08:58Z | http://arxiv.org/pdf/2406.09330v1 | http://arxiv.org/abs/2406.09330v1 | Somin Wadhwa, Adit Krishnan, Runhui Wang, Byron C. Wallace, Chris Kong | 2024-06-13T17:08:58Z | cs.CL | Entity matching is the task of linking records from different sources that
refer to the same real-world entity. Past work has primarily treated entity
linking as a standard supervised learning problem. However, supervised entity
matching models often do not generalize well to new data, and collecting
exhaustive labeled training data is often cost prohibitive. Further, recent
efforts have adopted LLMs for this task in few/zero-shot settings, exploiting
their general knowledge. But LLMs are prohibitively expensive for performing
inference at scale for real-world entity matching tasks.
As an efficient alternative, we re-cast entity matching as a conditional
generation task as opposed to binary classification. This enables us to
"distill" LLM reasoning into smaller entity matching models via natural
language explanations. This approach achieves strong performance, especially on
out-of-domain generalization tests (10.85% F-1) where standalone generative
methods struggle. We perform ablations that highlight the importance of
explanations, both for performance and model robustness. | cs.CL | Learning from Natural Language Explanations for Generalizable Entity Matching |
2024-06-13T17:02:32Z | http://arxiv.org/pdf/2406.09325v1 | http://arxiv.org/abs/2406.09325v1 | Tomer Ashuach, Martin Tutek, Yonatan Belinkov | 2024-06-13T17:02:32Z | cs.CL, I.2.7 | Large language models (LLMs) risk inadvertently memorizing and divulging
sensitive or personally identifiable information (PII) seen in training data,
causing privacy concerns. Current approaches to address this issue involve
costly dataset scrubbing, or model filtering through unlearning and model
editing, which can be bypassed through extraction attacks. We propose REVS, a
novel model editing method for unlearning sensitive information from LLMs. REVS
identifies and modifies a small subset of neurons relevant for each piece of
sensitive information. By projecting these neurons to the vocabulary space
(unembedding), we pinpoint the components driving its generation. We then
compute a model edit based on the pseudo-inverse of the unembedding matrix, and
apply it to de-promote generation of the targeted sensitive data. To adequately
evaluate our method on truly sensitive information, we curate two datasets: an
email dataset inherently memorized by GPT-J, and a synthetic social security
number dataset that we tune the model to memorize. Compared to other
state-of-the-art model editing methods, REVS demonstrates superior performance
in both eliminating sensitive information and robustness to extraction attacks,
while retaining integrity of the underlying model. The code and a demo notebook
are available at https://technion-cs-nlp.github.io/REVS. | cs.CL | REVS: Unlearning Sensitive Information in Language Models via Rank Editing in the Vocabulary Space |
2024-06-13T17:01:40Z | http://arxiv.org/pdf/2406.09324v1 | http://arxiv.org/abs/2406.09324v1 | Zhao Xu, Fan Liu, Hao Liu | 2024-06-13T17:01:40Z | cs.CR, cs.AI, cs.CL | Although Large Language Models (LLMs) have demonstrated significant
capabilities in executing complex tasks in a zero-shot manner, they are
susceptible to jailbreak attacks and can be manipulated to produce harmful
outputs. Recently, a growing body of research has categorized jailbreak attacks
into token-level and prompt-level attacks. However, previous work primarily
overlooks the diverse key factors of jailbreak attacks, with most studies
concentrating on LLM vulnerabilities and lacking exploration of
defense-enhanced LLMs. To address these issues, we evaluate the impact of
various attack settings on LLM performance and provide a baseline benchmark for
jailbreak attacks, encouraging the adoption of a standardized evaluation
framework. Specifically, we evaluate the eight key factors of implementing
jailbreak attacks on LLMs from both target-level and attack-level perspectives.
We further conduct seven representative jailbreak attacks on six defense
methods across two widely used datasets, encompassing approximately 320
experiments with about 50,000 GPU hours on A800-80G. Our experimental results
highlight the need for standardized benchmarking to evaluate these attacks on
defense-enhanced LLMs. Our code is available at
https://github.com/usail-hkust/Bag_of_Tricks_for_LLM_Jailbreaking. | cs.CR | Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs |
2024-06-13T16:59:43Z | http://arxiv.org/pdf/2406.09321v1 | http://arxiv.org/abs/2406.09321v1 | Delong Ran, Jinyuan Liu, Yichen Gong, Jingyi Zheng, Xinlei He, Tianshuo Cong, Anyu Wang | 2024-06-13T16:59:43Z | cs.CR, cs.AI, cs.CL | Jailbreak attacks aim to induce Large Language Models (LLMs) to generate
harmful responses for forbidden instructions, presenting severe misuse threats
to LLMs. Up to now, research into jailbreak attacks and defenses is emerging,
however, there is (surprisingly) no consensus on how to evaluate whether a
jailbreak attempt is successful. In other words, the methods to assess the
harmfulness of an LLM's response are varied, such as manual annotation or
prompting GPT-4 in specific ways. Each approach has its own set of strengths
and weaknesses, impacting their alignment with human values, as well as the
time and financial cost. This diversity in evaluation presents challenges for
researchers in choosing suitable evaluation methods and conducting fair
comparisons across different jailbreak attacks and defenses. In this paper, we
conduct a comprehensive analysis of jailbreak evaluation methodologies, drawing
from nearly ninety jailbreak research released between May 2023 and April 2024.
Our study introduces a systematic taxonomy of jailbreak evaluators, offering
in-depth insights into their strengths and weaknesses, along with the current
status of their adaptation. Moreover, to facilitate subsequent research, we
propose JailbreakEval, a user-friendly toolkit focusing on the evaluation of
jailbreak attempts. It includes various well-known evaluators out-of-the-box,
so that users can obtain evaluation results with only a single command.
JailbreakEval also allows users to customize their own evaluation workflow in a
unified framework with the ease of development and comparison. In summary, we
regard JailbreakEval to be a catalyst that simplifies the evaluation process in
jailbreak research and fosters an inclusive standard for jailbreak evaluation
within the community. | cs.CR | JailbreakEval: An Integrated Toolkit for Evaluating Jailbreak Attempts Against Large Language Models |
2024-06-13T16:58:02Z | http://arxiv.org/pdf/2406.09320v2 | http://arxiv.org/abs/2406.09320v2 | Nimol Thuon | 2024-06-16T19:08:34Z | cs.IR, cs.CL | The search engine process is crucial for document content retrieval. For
Khmer documents, an effective tool is needed to extract essential keywords and
facilitate accurate searches. Despite the daily generation of significant Khmer
content, Cambodians struggle to find necessary documents due to the lack of an
effective semantic searching tool. Even Google does not deliver high accuracy
for Khmer content. Semantic search engines improve search results by employing
advanced algorithms to understand various content types. With the rise in Khmer
digital content such as reports, articles, and social media feedback enhanced
search capabilities are essential. This research proposes the first Khmer
Semantic Search Engine (KSE), designed to enhance traditional Khmer search
methods. Utilizing semantic matching techniques and formally annotated semantic
content, our tool extracts meaningful keywords from user queries, performs
precise matching, and provides the best matching offline documents and online
URLs. We propose three semantic search frameworks: semantic search based on a
keyword dictionary, semantic search based on ontology, and semantic search
based on ranking. Additionally, we developed tools for data preparation,
including document addition and manual keyword extraction. To evaluate
performance, we created a ground truth dataset and addressed issues related to
searching and semantic search. Our findings demonstrate that understanding
search term semantics can lead to significantly more accurate results. | cs.IR | Khmer Semantic Search Engine (KSE): Digital Information Access and Document Retrieval |
2024-06-13T16:42:06Z | http://arxiv.org/pdf/2406.09308v1 | http://arxiv.org/abs/2406.09308v1 | Wilfried Bounsi, Borja Ibarz, Andrew Dudzik, Jessica B. Hamrick, Larisa Markeeva, Alex Vitvitskyi, Razvan Pascanu, Petar Veličković | 2024-06-13T16:42:06Z | cs.CL, cs.LG | Transformers have revolutionized machine learning with their simple yet
effective architecture. Pre-training Transformers on massive text datasets from
the Internet has led to unmatched generalization for natural language
understanding (NLU) tasks. However, such language models remain fragile when
tasked with algorithmic forms of reasoning, where computations must be precise
and robust. To address this limitation, we propose a novel approach that
combines the Transformer's language understanding with the robustness of graph
neural network (GNN)-based neural algorithmic reasoners (NARs). Such NARs
proved effective as generic solvers for algorithmic tasks, when specified in
graph form. To make their embeddings accessible to a Transformer, we propose a
hybrid architecture with a two-phase training procedure, allowing the tokens in
the language model to cross-attend to the node embeddings from the NAR. We
evaluate our resulting TransNAR model on CLRS-Text, the text-based version of
the CLRS-30 benchmark, and demonstrate significant gains over Transformer-only
models for algorithmic reasoning, both in and out of distribution. | cs.CL | Transformers meet Neural Algorithmic Reasoners |
2024-06-13T16:30:14Z | http://arxiv.org/pdf/2406.09295v2 | http://arxiv.org/abs/2406.09295v2 | Yuhang Wu, Wenmeng Yu, Yean Cheng, Yan Wang, Xiaohan Zhang, Jiazheng Xu, Ming Ding, Yuxiao Dong | 2024-06-14T02:14:49Z | cs.CL, cs.CV | Evaluating the alignment capabilities of large Vision-Language Models (VLMs)
is essential for determining their effectiveness as helpful assistants.
However, existing benchmarks primarily focus on basic abilities using nonverbal
methods, such as yes-no and multiple-choice questions. In this paper, we
address this gap by introducing AlignMMBench, a comprehensive alignment
benchmark specifically designed for emerging Chinese VLMs. This benchmark is
meticulously curated from real-world scenarios and Chinese Internet sources,
encompassing thirteen specific tasks across three categories, and includes both
single-turn and multi-turn dialogue scenarios. Incorporating a prompt rewrite
strategy, AlignMMBench encompasses 1,054 images and 4,978 question-answer
pairs. To facilitate the evaluation pipeline, we propose CritiqueVLM, a
rule-calibrated evaluator that exceeds GPT-4's evaluation ability. Finally, we
report the performance of representative VLMs on AlignMMBench, offering
insights into the capabilities and limitations of different VLM architectures.
All evaluation codes and data are available on https://alignmmbench.github.io. | cs.CL | AlignMMBench: Evaluating Chinese Multimodal Alignment in Large Vision-Language Models |
2024-06-13T16:27:56Z | http://arxiv.org/pdf/2406.09290v1 | http://arxiv.org/abs/2406.09290v1 | Martina Valente, Fabio Brugnara, Giovanni Morrone, Enrico Zovato, Leonardo Badino | 2024-06-13T16:27:56Z | eess.AS, cs.CL, cs.SD | This paper addresses spoken language identification (SLI) and speech
recognition of multilingual broadcast and institutional speech, real
application scenarios that have been rarely addressed in the SLI literature.
Observing that in these domains language changes are mostly associated with
speaker changes, we propose a cascaded system consisting of speaker diarization
and language identification and compare it with more traditional language
identification and language diarization systems. Results show that the proposed
system often achieves lower language classification and language diarization
error rates (up to 10% relative language diarization error reduction and 60%
relative language confusion reduction) and leads to lower WERs on multilingual
test sets (more than 8% relative WER reduction), while at the same time does
not negatively affect speech recognition on monolingual audio (with an absolute
WER increase between 0.1% and 0.7% w.r.t. monolingual ASR). | eess.AS | Exploring Spoken Language Identification Strategies for Automatic Transcription of Multilingual Broadcast and Institutional Speech |
2024-06-13T16:26:47Z | http://arxiv.org/pdf/2406.09289v1 | http://arxiv.org/abs/2406.09289v1 | Sarah Ball, Frauke Kreuter, Nina Rimsky | 2024-06-13T16:26:47Z | cs.CL, cs.AI, cs.LG | Conversational Large Language Models are trained to refuse to answer harmful
questions. However, emergent jailbreaking techniques can still elicit unsafe
outputs, presenting an ongoing challenge for model alignment. To better
understand how different jailbreak types circumvent safeguards, this paper
analyses model activations on different jailbreak inputs. We find that it is
possible to extract a jailbreak vector from a single class of jailbreaks that
works to mitigate jailbreak effectiveness from other classes. This may indicate
that different kinds of effective jailbreaks operate via similar internal
mechanisms. We investigate a potential common mechanism of harmfulness feature
suppression, and provide evidence for its existence by looking at the
harmfulness vector component. These findings offer actionable insights for
developing more robust jailbreak countermeasures and lay the groundwork for a
deeper, mechanistic understanding of jailbreak dynamics in language models. | cs.CL | Understanding Jailbreak Success: A Study of Latent Space Dynamics in Large Language Models |
2024-06-13T16:22:37Z | http://arxiv.org/pdf/2406.09282v1 | http://arxiv.org/abs/2406.09282v1 | Jinchuan Tian, Yifan Peng, William Chen, Kwanghee Choi, Karen Livescu, Shinji Watanabe | 2024-06-13T16:22:37Z | cs.CL, cs.SD, eess.AS | The Open Whisper-style Speech Model (OWSM) series was introduced to achieve
full transparency in building advanced speech-to-text (S2T) foundation models.
To this end, OWSM models are trained on 25 public speech datasets, which are
heterogeneous in multiple ways. In this study, we advance the OWSM series by
introducing OWSM v3.2, which improves on prior models by investigating and
addressing the impacts of this data heterogeneity. Our study begins with a
detailed analysis of each dataset, from which we derive two key strategies:
data filtering with proxy task to enhance data quality, and the incorporation
of punctuation and true-casing using an open large language model (LLM). With
all other configurations staying the same, OWSM v3.2 improves performance over
the OWSM v3.1 baseline while using 15% less training data. | cs.CL | On the Effects of Heterogeneous Data Sources on Speech-to-Text Foundation Models |
2024-06-13T16:20:05Z | http://arxiv.org/pdf/2406.09490v1 | http://arxiv.org/abs/2406.09490v1 | Emily Silcock, Abhishek Arora, Luca D'Amico-Wong, Melissa Dell | 2024-06-13T16:20:05Z | cs.CL, econ.GN, q-fin.EC | In the U.S. historically, local newspapers drew their content largely from
newswires like the Associated Press. Historians argue that newswires played a
pivotal role in creating a national identity and shared understanding of the
world, but there is no comprehensive archive of the content sent over
newswires. We reconstruct such an archive by applying a customized deep
learning pipeline to hundreds of terabytes of raw image scans from thousands of
local newspapers. The resulting dataset contains 2.7 million unique public
domain U.S. newswire articles, written between 1878 and 1977. Locations in
these articles are georeferenced, topics are tagged using customized neural
topic classification, named entities are recognized, and individuals are
disambiguated to Wikipedia using a novel entity disambiguation model. To
construct the Newswire dataset, we first recognize newspaper layouts and
transcribe around 138 millions structured article texts from raw image scans.
We then use a customized neural bi-encoder model to de-duplicate reproduced
articles, in the presence of considerable abridgement and noise, quantifying
how widely each article was reproduced. A text classifier is used to ensure
that we only include newswire articles, which historically are in the public
domain. The structured data that accompany the texts provide rich information
about the who (disambiguated individuals), what (topics), and where
(georeferencing) of the news that millions of Americans read over the course of
a century. We also include Library of Congress metadata information about the
newspapers that ran the articles on their front pages. The Newswire dataset is
useful both for large language modeling - expanding training data beyond what
is available from modern web texts - and for studying a diversity of questions
in computational linguistics, social science, and the digital humanities. | cs.CL | Newswire: A Large-Scale Structured Database of a Century of Historical News |
2024-06-13T16:17:21Z | http://arxiv.org/pdf/2406.09279v1 | http://arxiv.org/abs/2406.09279v1 | Hamish Ivison, Yizhong Wang, Jiacheng Liu, Zeqiu Wu, Valentina Pyatkin, Nathan Lambert, Noah A. Smith, Yejin Choi, Hannaneh Hajishirzi | 2024-06-13T16:17:21Z | cs.CL | Learning from preference feedback has emerged as an essential step for
improving the generation quality and performance of modern language models
(LMs). Despite its widespread use, the way preference-based learning is applied
varies wildly, with differing data, learning algorithms, and evaluations used,
making disentangling the impact of each aspect difficult. In this work, we
identify four core aspects of preference-based learning: preference data,
learning algorithm, reward model, and policy training prompts, systematically
investigate the impact of these components on downstream model performance, and
suggest a recipe for strong learning for preference feedback. Our findings
indicate that all aspects are important for performance, with better preference
data leading to the largest improvements, followed by the choice of learning
algorithm, the use of improved reward models, and finally the use of additional
unlabeled prompts for policy training. Notably, PPO outperforms DPO by up to
2.5% in math and 1.2% in general domains. High-quality preference data leads to
improvements of up to 8% in instruction following and truthfulness. Despite
significant gains of up to 5% in mathematical evaluation when scaling up reward
models, we surprisingly observe marginal improvements in other categories.
We publicly release the code used for training
(https://github.com/hamishivi/EasyLM) and evaluating
(https://github.com/allenai/open-instruct) our models, along with the models
and datasets themselves
(https://huggingface.co/collections/allenai/tulu-v25-suite-66676520fd578080e126f618). | cs.CL | Unpacking DPO and PPO: Disentangling Best Practices for Learning from Preference Feedback |
2024-06-13T16:15:53Z | http://arxiv.org/pdf/2406.09277v1 | http://arxiv.org/abs/2406.09277v1 | Waris Quamer, Ricardo Gutierrez-Osuna | 2024-06-13T16:15:53Z | eess.AS, cs.CL, cs.LG | Speaker anonymization aims to conceal cues to speaker identity while
preserving linguistic content. Current machine learning based approaches
require substantial computational resources, hindering real-time streaming
applications. To address these concerns, we propose a streaming model that
achieves speaker anonymization with low latency. The system is trained in an
end-to-end autoencoder fashion using a lightweight content encoder that
extracts HuBERT-like information, a pretrained speaker encoder that extract
speaker identity, and a variance encoder that injects pitch and energy
information. These three disentangled representations are fed to a decoder that
resynthesizes the speech signal. We present evaluation results from two
implementations of our system, a full model that achieves a latency of 230ms,
and a lite version (0.1x in size) that further reduces latency to 66ms while
maintaining state-of-the-art performance in naturalness, intelligibility, and
privacy preservation. | eess.AS | End-to-end Streaming model for Low-Latency Speech Anonymization |
2024-06-13T16:04:11Z | http://arxiv.org/pdf/2406.09265v1 | http://arxiv.org/abs/2406.09265v1 | Weixuan Wang, Barry Haddow, Wei Peng, Alexandra Birch | 2024-06-13T16:04:11Z | cs.CL | Multilingual large language models (LLMs) have greatly increased the ceiling
of performance on non-English tasks. However the mechanisms behind
multilingualism in these LLMs are poorly understood. Of particular interest is
the degree to which internal representations are shared between languages.
Recent work on neuron analysis of LLMs has focused on the monolingual case, and
the limited work on the multilingual case has not considered the interaction
between tasks and linguistic representations. In our work, we investigate how
neuron activation is shared across languages by categorizing neurons into four
distinct groups according to their responses across different languages for a
particular input: all-shared, partial-shared, specific, and non-activated. This
categorization is combined with a study of neuron attribution, i.e. the
importance of a neuron w.r.t an output. Our analysis reveals the following
insights: (i) the linguistic sharing patterns are strongly affected by the type
of task, but neuron behaviour changes across different inputs even for the same
task; (ii) all-shared neurons play a key role in generating correct responses;
(iii) boosting multilingual alignment by increasing all-shared neurons can
enhance accuracy on multilingual tasks. The code is available at
https://github.com/weixuan-wang123/multilingual-neurons. | cs.CL | Sharing Matters: Analysing Neurons Across Languages and Tasks in LLMs |
2024-06-13T16:03:25Z | http://arxiv.org/pdf/2406.09264v2 | http://arxiv.org/abs/2406.09264v2 | Hua Shen, Tiffany Knearem, Reshmi Ghosh, Kenan Alkiek, Kundan Krishna, Yachuan Liu, Ziqiao Ma, Savvas Petridis, Yi-Hao Peng, Li Qiwei, Sushrita Rakshit, Chenglei Si, Yutong Xie, Jeffrey P. Bigham, Frank Bentley, Joyce Chai, Zachary Lipton, Qiaozhu Mei, Rada Mihalcea, Michael Terry, Diyi Yang, Meredith Ringel Morris, Paul Resnick, David Jurgens | 2024-06-17T16:58:35Z | cs.HC, cs.AI, cs.CL | Recent advancements in general-purpose AI have highlighted the importance of
guiding AI systems towards the intended goals, ethical principles, and values
of individuals and groups, a concept broadly recognized as alignment. However,
the lack of clarified definitions and scopes of human-AI alignment poses a
significant obstacle, hampering collaborative efforts across research domains
to achieve this alignment. In particular, ML- and philosophy-oriented alignment
research often views AI alignment as a static, unidirectional process (i.e.,
aiming to ensure that AI systems' objectives match humans) rather than an
ongoing, mutual alignment problem [429]. This perspective largely neglects the
long-term interaction and dynamic changes of alignment. To understand these
gaps, we introduce a systematic review of over 400 papers published between
2019 and January 2024, spanning multiple domains such as Human-Computer
Interaction (HCI), Natural Language Processing (NLP), Machine Learning (ML),
and others. We characterize, define and scope human-AI alignment. From this, we
present a conceptual framework of "Bidirectional Human-AI Alignment" to
organize the literature from a human-centered perspective. This framework
encompasses both 1) conventional studies of aligning AI to humans that ensures
AI produces the intended outcomes determined by humans, and 2) a proposed
concept of aligning humans to AI, which aims to help individuals and society
adjust to AI advancements both cognitively and behaviorally. Additionally, we
articulate the key findings derived from literature analysis, including
discussions about human values, interaction techniques, and evaluations. To
pave the way for future studies, we envision three key challenges for future
directions and propose examples of potential future solutions. | cs.HC | Towards Bidirectional Human-AI Alignment: A Systematic Review for Clarifications, Framework, and Future Directions |
2024-06-13T15:06:11Z | http://arxiv.org/pdf/2406.09206v1 | http://arxiv.org/abs/2406.09206v1 | Christopher Schröder, Gerhard Heyer | 2024-06-13T15:06:11Z | cs.CL, cs.AI, cs.LG | Active learning is an iterative labeling process that is used to obtain a
small labeled subset, despite the absence of labeled data, thereby enabling to
train a model for supervised tasks such as text classification. While active
learning has made considerable progress in recent years due to improvements
provided by pre-trained language models, there is untapped potential in the
often neglected unlabeled portion of the data, although it is available in
considerably larger quantities than the usually small set of labeled data. Here
we investigate how self-training, a semi-supervised approach where a model is
used to obtain pseudo-labels from the unlabeled data, can be used to improve
the efficiency of active learning for text classification. Starting with an
extensive reproduction of four previous self-training approaches, some of which
are evaluated for the first time in the context of active learning or natural
language processing, we devise HAST, a new and effective self-training
strategy, which is evaluated on four text classification benchmarks, on which
it outperforms the reproduced self-training approaches and reaches
classification results comparable to previous experiments for three out of four
datasets, using only 25% of the data. | cs.CL | Self-Training for Sample-Efficient Active Learning for Text Classification with Pre-Trained Language Models |
2024-06-13T15:03:46Z | http://arxiv.org/pdf/2406.09205v1 | http://arxiv.org/abs/2406.09205v1 | Hieu Tran, Zonghai Yao, Lingxi Li, Hong Yu | 2024-06-13T15:03:46Z | cs.CL, cs.AI | Content generation conditioning on users's readability is an important
application for personalization. In an era of large language models (LLMs),
readability-controlled text generation based on LLMs has become increasingly
important. This paper introduces a novel methodology called
"Readability-Controlled Instruction Learning (ReadCtrl)," which aims to
instruction-tune LLMs to tailor users' readability levels. Unlike the
traditional methods, which primarily focused on categorical readability
adjustments typically classified as high, medium, and low or expert and
layperson levels with limited success, ReadCtrl introduces a dynamic framework
that enables LLMs to generate content at various (near continuous level)
complexity levels, thereby enhancing their versatility across different
applications. Our results show that the ReadCtrl-Mistral-7B models
significantly outperformed strong baseline models such as GPT-4 and Claude-3,
with a win rate of 52.1%:35.7% against GPT-4 in human evaluations. Furthermore,
Read-Ctrl has shown significant improvements in automatic evaluations, as
evidenced by better readability metrics (e.g., FOG, FKGL) and generation
quality metrics (e.g., BLEU, SARI, SummaC-Factuality, UniEval-Consistency and
Coherence). These results underscore Read-Ctrl's effectiveness and tenacity in
producing high-quality, contextually appropriate outputs that closely align
with targeted readability levels, marking a significant advancement in
personalized content generation using LLMs. | cs.CL | ReadCtrl: Personalizing text generation with readability-controlled instruction learning |
2024-06-13T14:59:45Z | http://arxiv.org/pdf/2406.09202v1 | http://arxiv.org/abs/2406.09202v1 | Chihiro Taguchi, David Chiang | 2024-06-13T14:59:45Z | cs.CL, cs.AI | We investigate what linguistic factors affect the performance of Automatic
Speech Recognition (ASR) models. We hypothesize that orthographic and
phonological complexities both degrade accuracy. To examine this, we fine-tune
the multilingual self-supervised pretrained model Wav2Vec2-XLSR-53 on 25
languages with 15 writing systems, and we compare their ASR accuracy, number of
graphemes, unigram grapheme entropy, logographicity (how much
word/morpheme-level information is encoded in the writing system), and number
of phonemes. The results demonstrate that orthographic complexities
significantly correlate with low ASR accuracy, while phonological complexity
shows no significant correlation. | cs.CL | Language Complexity and Speech Recognition Accuracy: Orthographic Complexity Hurts, Phonological Complexity Doesn't |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.