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
|
|---|---|---|---|---|---|---|---|---|
2024-02-16T16:15:01Z | http://arxiv.org/pdf/2402.10790v2 | http://arxiv.org/abs/2402.10790v2 | Yuri Kuratov, Aydar Bulatov, Petr Anokhin, Dmitry Sorokin, Artyom Sorokin, Mikhail Burtsev | 2024-02-21T03:07:42Z | cs.CL, cs.AI, cs.LG | This paper addresses the challenge of processing long documents using
generative transformer models. To evaluate different approaches, we introduce
BABILong, a new benchmark designed to assess model capabilities in extracting
and processing distributed facts within extensive texts. Our evaluation, which
includes benchmarks for GPT-4 and RAG, reveals that common methods are
effective only for sequences up to $10^4$ elements. In contrast, fine-tuning
GPT-2 with recurrent memory augmentations enables it to handle tasks involving
up to $11\times 10^6$ elements. This achievement marks a substantial leap, as
it is by far the longest input processed by any neural network model to date,
demonstrating a significant improvement in the processing capabilities for long
sequences. | cs.CL | In Search of Needles in a 11M Haystack: Recurrent Memory Finds What LLMs Miss |
2024-02-16T16:10:38Z | http://arxiv.org/pdf/2402.10787v1 | http://arxiv.org/abs/2402.10787v1 | Xuan Shen, Zhenglun Kong, Changdi Yang, Zhaoyang Han, Lei Lu, Peiyan Dong, Cheng Lyu, Chih-hsiang Li, Xuehang Guo, Zhihao Shu, Wei Niu, Miriam Leeser, Pu Zhao, Yanzhi Wang | 2024-02-16T16:10:38Z | cs.LG, cs.AI, cs.CL | Despite the remarkable strides of Large Language Models (LLMs) in various
fields, the wide applications of LLMs on edge devices are limited due to their
massive parameters and computations. To address this, quantization is commonly
adopted to generate lightweight LLMs with efficient computations and fast
inference. However, Post-Training Quantization (PTQ) methods dramatically
degrade in quality when quantizing weights, activations, and KV cache together
to below 8 bits. Besides, many Quantization-Aware Training (QAT) works quantize
model weights, leaving the activations untouched, which do not fully exploit
the potential of quantization for inference acceleration on the edge. In this
paper, we propose EdgeQAT, the Entropy and Distribution Guided QAT for the
optimization of lightweight LLMs to achieve inference acceleration on Edge
devices. We first identify that the performance drop of quantization primarily
stems from the information distortion in quantized attention maps, demonstrated
by the different distributions in quantized query and key of the self-attention
mechanism. Then, the entropy and distribution guided QAT is proposed to
mitigate the information distortion. Moreover, we design a token
importance-aware adaptive method to dynamically quantize the tokens with
different bit widths for further optimization and acceleration. Our extensive
experiments verify the substantial improvements with our framework across
various datasets. Furthermore, we achieve an on-device speedup of up to 2.37x
compared with its FP16 counterparts across multiple edge devices, signaling a
groundbreaking advancement. | cs.LG | EdgeQAT: Entropy and Distribution Guided Quantization-Aware Training for the Acceleration of Lightweight LLMs on the Edge |
2024-02-16T16:02:33Z | http://arxiv.org/pdf/2402.10779v1 | http://arxiv.org/abs/2402.10779v1 | Mingchen Li, Chen Ling, Rui Zhang, Liang Zhao | 2024-02-16T16:02:33Z | cs.CL | Zero-shot link prediction (ZSLP) on knowledge graphs aims at automatically
identifying relations between given entities. Existing methods primarily employ
auxiliary information to predict tail entity given head entity and its
relation, yet face challenges due to the occasional unavailability of such
detailed information and the inherent simplicity of predicting tail entities
based on semantic similarities. Even though Large Language Models (LLMs) offer
a promising solution to predict unobserved relations between the head and tail
entity in a zero-shot manner, their performance is still restricted due to the
inability to leverage all the (exponentially many) paths' information between
two entities, which are critical in collectively indicating their relation
types. To address this, in this work, we introduce a Condensed Transition Graph
Framework for Zero-Shot Link Prediction (CTLP), which encodes all the paths'
information in linear time complexity to predict unseen relations between
entities, attaining both efficiency and information preservation. Specifically,
we design a condensed transition graph encoder with theoretical guarantees on
its coverage, expressiveness, and efficiency. It is learned by a transition
graph contrastive learning strategy. Subsequently, we design a soft instruction
tuning to learn and map the all-path embedding to the input of LLMs.
Experimental results show that our proposed CTLP method achieves
state-of-the-art performance on three standard ZSLP datasets | cs.CL | A Condensed Transition Graph Framework for Zero-shot Link Prediction with Large Language Models |
2024-02-16T15:54:24Z | http://arxiv.org/pdf/2402.10772v1 | http://arxiv.org/abs/2402.10772v1 | Hariram Veeramani, Surendrabikram Thapa, Usman Naseem | 2024-02-16T15:54:24Z | cs.CL | In the evolving landscape of Environmental, Social, and Corporate Governance
(ESG) impact assessment, the ML-ESG-2 shared task proposes identifying ESG
impact types. To address this challenge, we present a comprehensive system
leveraging ensemble learning techniques, capitalizing on early and late fusion
approaches. Our approach employs four distinct models: mBERT, FlauBERT-base,
ALBERT-base-v2, and a Multi-Layer Perceptron (MLP) incorporating Latent
Semantic Analysis (LSA) and Term Frequency-Inverse Document Frequency (TF-IDF)
features. Through extensive experimentation, we find that our early fusion
ensemble approach, featuring the integration of LSA, TF-IDF, mBERT,
FlauBERT-base, and ALBERT-base-v2, delivers the best performance. Our system
offers a comprehensive ESG impact type identification solution, contributing to
the responsible and sustainable decision-making processes vital in today's
financial and corporate governance landscape. | cs.CL | Enhancing ESG Impact Type Identification through Early Fusion and Multilingual Models |
2024-02-16T15:48:33Z | http://arxiv.org/pdf/2402.10770v3 | http://arxiv.org/abs/2402.10770v3 | Ehsan Doostmohammadi, Oskar Holmström, Marco Kuhlmann | 2024-07-02T11:46:09Z | cs.CL, cs.AI | Work on instruction-tuned Large Language Models (LLMs) has used automatic
methods based on text overlap and LLM judgments as cost-effective alternatives
to human evaluation. In this paper, we perform a meta-evaluation of such
methods and assess their reliability across a broad range of tasks. We observe
that while automatic evaluation methods can approximate human ratings under
specific conditions, their validity is highly context-dependent. Specifically,
the simple ROUGE-L metric correlates well with human ratings for short-answer
English tasks but is unreliable in free-form generation tasks and cross-lingual
transfer. The effectiveness of the more advanced method of using GPT-4 as a
judge diminishes significantly if reference answers are not included in the
prompt, which is the scenario where this method has the potential to provide
the most value compared to other metrics. Our findings enhance the
understanding of how automatic methods should be applied and interpreted when
developing and evaluating instruction-tuned LLMs. | cs.CL | How Reliable Are Automatic Evaluation Methods for Instruction-Tuned LLMs? |
2024-02-16T15:48:24Z | http://arxiv.org/pdf/2402.10769v1 | http://arxiv.org/abs/2402.10769v1 | Yongqi Li, Zhen Zhang, Wenjie Wang, Liqiang Nie, Wenjie Li, Tat-Seng Chua | 2024-02-16T15:48:24Z | cs.CL, cs.AI, cs.IR | Generative retrieval is a promising new paradigm in text retrieval that
generates identifier strings of relevant passages as the retrieval target. This
paradigm leverages powerful generative language models, distinct from
traditional sparse or dense retrieval methods. In this work, we identify a
viable direction to further enhance generative retrieval via distillation and
propose a feasible framework, named DGR. DGR utilizes sophisticated ranking
models, such as the cross-encoder, in a teacher role to supply a passage rank
list, which captures the varying relevance degrees of passages instead of
binary hard labels; subsequently, DGR employs a specially designed distilled
RankNet loss to optimize the generative retrieval model, considering the
passage rank order provided by the teacher model as labels. This framework only
requires an additional distillation step to enhance current generative
retrieval systems and does not add any burden to the inference stage. We
conduct experiments on four public datasets, and the results indicate that DGR
achieves state-of-the-art performance among the generative retrieval methods.
Additionally, DGR demonstrates exceptional robustness and generalizability with
various teacher models and distillation losses. | cs.CL | Distillation Enhanced Generative Retrieval |
2024-02-16T15:41:23Z | http://arxiv.org/pdf/2402.10767v1 | http://arxiv.org/abs/2402.10767v1 | Dhairya Dalal, Marco Valentino, André Freitas, Paul Buitelaar | 2024-02-16T15:41:23Z | cs.CL, cs.AI, I.2.7 | While Large Language Models (LLMs) have found success in real-world
applications, their underlying explanatory process is still poorly understood.
This paper proposes IBE-Eval, a framework inspired by philosophical accounts on
Inference to the Best Explanation (IBE) to advance the interpretation and
evaluation of LLMs' explanations. IBE-Eval estimates the plausibility of
natural language explanations through a combination of explicit logical and
linguistic features including: consistency, parsimony, coherence, and
uncertainty. Extensive experiments are conducted on Causal Question Answering
(CQA), where \textit{IBE-Eval} is tasked to select the most plausible causal
explanation amongst competing ones generated by LLMs (i.e., GPT 3.5 and Llama
2). The experiments reveal that IBE-Eval can successfully identify the best
explanation with up to 77\% accuracy ($\approx 27\%$ above random), improving
upon a GPT 3.5-as-a-Judge baseline ($\approx+17\%$) while being intrinsically
more efficient and interpretable. Additional analyses suggest that, despite
model-specific variances, LLM-generated explanations tend to conform to IBE
criteria and that IBE-Eval is significantly correlated with human judgment,
opening up opportunities for future development of automated explanation
verification tools. | cs.CL | Inference to the Best Explanation in Large Language Models |
2024-02-16T15:19:46Z | http://arxiv.org/pdf/2402.10753v1 | http://arxiv.org/abs/2402.10753v1 | Junjie Ye, Sixian Li, Guanyu Li, Caishuang Huang, Songyang Gao, Yilong Wu, Qi Zhang, Tao Gui, Xuanjing Huang | 2024-02-16T15:19:46Z | cs.CL, cs.AI | Tool learning is widely acknowledged as a foundational approach or deploying
large language models (LLMs) in real-world scenarios. While current research
primarily emphasizes leveraging tools to augment LLMs, it frequently neglects
emerging safety considerations tied to their application. To fill this gap, we
present $ToolSword$, a comprehensive framework dedicated to meticulously
investigating safety issues linked to LLMs in tool learning. Specifically,
ToolSword delineates six safety scenarios for LLMs in tool learning,
encompassing $malicious$ $queries$ and $jailbreak$ $attacks$ in the input
stage, $noisy$ $misdirection$ and $risky$ $cues$ in the execution stage, and
$harmful$ $feedback$ and $error$ $conflicts$ in the output stage. Experiments
conducted on 11 open-source and closed-source LLMs reveal enduring safety
challenges in tool learning, such as handling harmful queries, employing risky
tools, and delivering detrimental feedback, which even GPT-4 is susceptible to.
Moreover, we conduct further studies with the aim of fostering research on tool
learning safety. The data is released in
https://github.com/Junjie-Ye/ToolSword. | cs.CL | ToolSword: Unveiling Safety Issues of Large Language Models in Tool Learning Across Three Stages |
2024-02-16T15:01:24Z | http://arxiv.org/pdf/2402.10744v1 | http://arxiv.org/abs/2402.10744v1 | Pengcheng Jiang, Jiacheng Lin, Zifeng Wang, Jimeng Sun, Jiawei Han | 2024-02-16T15:01:24Z | cs.CL, cs.AI | The field of relation extraction (RE) is experiencing a notable shift towards
generative relation extraction (GRE), leveraging the capabilities of large
language models (LLMs). However, we discovered that traditional relation
extraction (RE) metrics like precision and recall fall short in evaluating GRE
methods. This shortfall arises because these metrics rely on exact matching
with human-annotated reference relations, while GRE methods often produce
diverse and semantically accurate relations that differ from the references. To
fill this gap, we introduce GenRES for a multi-dimensional assessment in terms
of the topic similarity, uniqueness, granularity, factualness, and completeness
of the GRE results. With GenRES, we empirically identified that (1)
precision/recall fails to justify the performance of GRE methods; (2)
human-annotated referential relations can be incomplete; (3) prompting LLMs
with a fixed set of relations or entities can cause hallucinations. Next, we
conducted a human evaluation of GRE methods that shows GenRES is consistent
with human preferences for RE quality. Last, we made a comprehensive evaluation
of fourteen leading LLMs using GenRES across document, bag, and sentence level
RE datasets, respectively, to set the benchmark for future research in GRE | cs.CL | GenRES: Rethinking Evaluation for Generative Relation Extraction in the Era of Large Language Models |
2024-02-16T14:59:55Z | http://arxiv.org/pdf/2402.10743v1 | http://arxiv.org/abs/2402.10743v1 | Hanqing Zhao, Yuehan Li | 2024-02-16T14:59:55Z | cs.CL | Entity and relationship extraction is a crucial component in natural language
processing tasks such as knowledge graph construction, question answering
system design, and semantic analysis. Most of the information of the Yishui
school of traditional Chinese Medicine (TCM) is stored in the form of
unstructured classical Chinese text. The key information extraction of TCM
texts plays an important role in mining and studying the academic schools of
TCM. In order to solve these problems efficiently using artificial intelligence
methods, this study constructs a word segmentation and entity relationship
extraction model based on conditional random fields under the framework of
natural language processing technology to identify and extract the entity
relationship of traditional Chinese medicine texts, and uses the common
weighting technology of TF-IDF information retrieval and data mining to extract
important key entity information in different ancient books. The dependency
syntactic parser based on neural network is used to analyze the grammatical
relationship between entities in each ancient book article, and it is
represented as a tree structure visualization, which lays the foundation for
the next construction of the knowledge graph of Yishui school and the use of
artificial intelligence methods to carry out the research of TCM academic
schools. | cs.CL | Construction of a Syntactic Analysis Map for Yi Shui School through Text Mining and Natural Language Processing Research |
2024-02-16T14:55:33Z | http://arxiv.org/pdf/2402.10738v2 | http://arxiv.org/abs/2402.10738v2 | Yinpeng Liu, Jiawei Liu, Xiang Shi, Qikai Cheng, Yong Huang, Wei Lu | 2024-06-16T13:26:10Z | cs.CL | Demonstration ordering, which is an important strategy for in-context
learning (ICL), can significantly affects the performance of large language
models (LLMs). However, most of the current approaches of ordering require high
computational costs to introduce the priori knowledge. In this paper, inspired
by the human learning process, we propose a simple but effective demonstration
ordering method for ICL, named the few-shot In-Context Curriculum Learning
(ICCL). The ICCL implies gradually increasing the complexity of prompt
demonstrations during the inference process. The difficulty can be assessed by
human experts or LLMs-driven metrics, such as perplexity. Then we design
extensive experiments to discuss the effectiveness of the ICCL at both
corpus-level and instance-level. Moreover, we also investigate the formation
mechanism of LLM's ICCL capability. Experimental results demonstrate that ICCL,
developed during the instruction-tuning stage, is effective for representative
open-source LLMs. To facilitate further research and applications by other
scholars, we make the code publicly available. | cs.CL | Let's Learn Step by Step: Enhancing In-Context Learning Ability with Curriculum Learning |
2024-02-16T14:52:05Z | http://arxiv.org/pdf/2402.10735v2 | http://arxiv.org/abs/2402.10735v2 | John Dougrez-Lewis, Mahmud Elahi Akhter, Yulan He, Maria Liakata | 2024-03-20T19:14:54Z | cs.CL | The reasoning capabilities of LLMs are currently hotly debated. We examine
the issue from the perspective of claim/rumour verification. We propose the
first logical reasoning framework designed to break down any claim or rumour
paired with evidence into the atomic reasoning steps necessary for
verification. Based on our framework, we curate two annotated collections of
such claim/evidence pairs: a synthetic dataset from Wikipedia and a real-world
set stemming from rumours circulating on Twitter. We use them to evaluate the
reasoning capabilities of GPT-3.5-Turbo and GPT-4 (hereinafter referred to as
ChatGPT) within the context of our framework, providing a thorough analysis.
Our results show that ChatGPT struggles in abductive reasoning, although this
can be somewhat mitigated by using manual Chain of Thought (CoT) as opposed to
Zero-Shot (ZS) and ZS CoT approaches. Our study contributes to the growing body
of research suggesting that ChatGPT's reasoning processes are unlikely to
mirror human-like reasoning, and that LLMs need to be more rigorously evaluated
to distinguish between hype and actual capabilities, especially in high-stakes
real-world tasks such as claim verification. | cs.CL | Assessing the Reasoning Abilities of ChatGPT in the Context of Claim Verification |
2024-02-16T14:44:30Z | http://arxiv.org/pdf/2404.04261v1 | http://arxiv.org/abs/2404.04261v1 | Nouar AlDahoul, Talal Rahwan, Yasir Zaki | 2024-02-16T14:44:30Z | cs.CL, cs.AI | A quarter of US adults regularly get their news from YouTube. Yet, despite
the massive political content available on the platform, to date no classifier
has been proposed to identify the political leaning of YouTube videos. To fill
this gap, we propose a novel classifier based on Bert -- a language model from
Google -- to classify YouTube videos merely based on their titles into six
categories, namely: Far Left, Left, Center, Anti-Woke, Right, and Far Right. We
used a public dataset of 10 million YouTube video titles (under various
categories) to train and validate the proposed classifier. We compare the
classifier against several alternatives that we trained on the same dataset,
revealing that our classifier achieves the highest accuracy (75%) and the
highest F1 score (77%). To further validate the classification performance, we
collect videos from YouTube channels of numerous prominent news agencies, such
as Fox News and New York Times, which have widely known political leanings, and
apply our classifier to their video titles. For the vast majority of cases, the
predicted political leaning matches that of the news agency. | cs.CL | A Novel BERT-based Classifier to Detect Political Leaning of YouTube Videos based on their Titles |
2024-02-16T14:23:55Z | http://arxiv.org/pdf/2402.10992v1 | http://arxiv.org/abs/2402.10992v1 | Holger Lyre | 2024-02-16T14:23:55Z | cs.CL, cs.AI | Do LLMs understand the meaning of the texts they generate? Do they possess a
semantic grounding? And how could we understand whether and what they
understand? I start the paper with the observation that we have recently
witnessed a generative turn in AI, since generative models, including LLMs, are
key for self-supervised learning. To assess the question of semantic grounding,
I distinguish and discuss five methodological ways. The most promising way is
to apply core assumptions of theories of meaning in philosophy of mind and
language to LLMs. Grounding proves to be a gradual affair with a
three-dimensional distinction between functional, social and causal grounding.
LLMs show basic evidence in all three dimensions. A strong argument is that
LLMs develop world models. Hence, LLMs are neither stochastic parrots nor
semantic zombies, but already understand the language they generate, at least
in an elementary sense. | cs.CL | "Understanding AI": Semantic Grounding in Large Language Models |
2024-02-16T14:15:15Z | http://arxiv.org/pdf/2402.10712v2 | http://arxiv.org/abs/2402.10712v2 | Atsuki Yamaguchi, Aline Villavicencio, Nikolaos Aletras | 2024-06-17T12:00:02Z | cs.CL, cs.AI | The development of state-of-the-art generative large language models (LLMs)
disproportionately relies on English-centric tokenizers, vocabulary and
pre-training data. Despite the fact that some LLMs have multilingual
capabilities, recent studies have shown that their inference efficiency
deteriorates when generating text in languages other than English. This results
in increased inference time and costs. Cross-lingual vocabulary adaptation
(CVA) methods have been proposed for adapting models to a target language
aiming to improve downstream performance. However, the effectiveness of these
methods on increasing inference efficiency of generative LLMs has yet to be
explored. In this paper, we perform an empirical study of five CVA methods on
four generative LLMs (including monolingual and multilingual models) across
four typologically-diverse languages and four natural language understanding
tasks. We find that CVA substantially contributes to LLM inference speedups of
up to 271.5\%. We also show that adapting LLMs that have been pre-trained on
more balanced multilingual data results in downstream performance comparable to
the original models. | cs.CL | An Empirical Study on Cross-lingual Vocabulary Adaptation for Efficient Language Model Inference |
2024-02-16T14:00:56Z | http://arxiv.org/pdf/2402.10699v1 | http://arxiv.org/abs/2402.10699v1 | Hongbin Na, Zimu Wang, Mieradilijiang Maimaiti, Tong Chen, Wei Wang, Tao Shen, Ling Chen | 2024-02-16T14:00:56Z | cs.CL | Large language models (LLMs) have demonstrated promising potential in various
downstream tasks, including machine translation. However, prior work on
LLM-based machine translation has mainly focused on better utilizing training
data, demonstrations, or pre-defined and universal knowledge to improve
performance, with a lack of consideration of decision-making like human
translators. In this paper, we incorporate Thinker with the Drift-Diffusion
Model (Thinker-DDM) to address this issue. We then redefine the Drift-Diffusion
process to emulate human translators' dynamic decision-making under constrained
resources. We conduct extensive experiments under the high-resource,
low-resource, and commonsense translation settings using the WMT22 and CommonMT
datasets, in which Thinker-DDM outperforms baselines in the first two
scenarios. We also perform additional analysis and evaluation on commonsense
translation to illustrate the high effectiveness and efficacy of the proposed
method. | cs.CL | Rethinking Human-like Translation Strategy: Integrating Drift-Diffusion Model with Large Language Models for Machine Translation |
2024-02-16T13:53:26Z | http://arxiv.org/pdf/2402.10693v3 | http://arxiv.org/abs/2402.10693v3 | Florian Le Bronnec, Alexandre Verine, Benjamin Negrevergne, Yann Chevaleyre, Alexandre Allauzen | 2024-06-04T11:33:27Z | cs.CL, cs.LG | We introduce a novel evaluation framework for Large Language Models (LLMs)
such as \textsc{Llama-2} and \textsc{Mistral}, focusing on importing Precision
and Recall metrics from image generation to text generation. This approach
allows for a nuanced assessment of the quality and diversity of generated text
without the need for aligned corpora. By conducting a comprehensive evaluation
of state-of-the-art language models, the study reveals new insights into their
performance on open-ended generation tasks, which are not adequately captured
by traditional benchmarks. The findings highlight a trade-off between the
quality and diversity of generated samples, particularly when models are
fine-tuned on instruction dataset or with human feedback. This work extends the
toolkit for distribution-based NLP evaluation, offering insights into the
practical capabilities and challenges that current LLMs face in generating
diverse and high-quality text. We release our code and data. | cs.CL | Exploring Precision and Recall to assess the quality and diversity of LLMs |
2024-02-16T13:48:06Z | http://arxiv.org/pdf/2402.10691v2 | http://arxiv.org/abs/2402.10691v2 | Xianzhen Luo, Qingfu Zhu, Zhiming Zhang, Libo Qin, Xuanyu Zhang, Qing Yang, Dongliang Xu, Wanxiang Che | 2024-06-16T12:29:50Z | cs.CL | Program of Thoughts (PoT) is an approach characterized by its executable
intermediate steps, which ensure the accuracy of the logical calculations in
the reasoning process. Currently, PoT primarily uses Python. However, relying
solely on a single language may result in suboptimal solutions and overlook the
potential benefits of other programming languages. In this paper, we conduct
comprehensive experiments on the programming languages used in PoT and find
that no single language consistently delivers optimal performance across all
tasks and models. The effectiveness of each language varies depending on the
specific scenarios. Inspired by this, we propose a task and model agnostic
approach called MultiPoT, which harnesses strength and diversity from various
languages. Experimental results reveal that it significantly outperforms Python
Self-Consistency. Furthermore, it achieves comparable or superior performance
compared to the best monolingual PoT in almost all tasks across all models. In
particular, MultiPoT achieves more than 4.6% improvement on average on ChatGPT
(gpt-3.5-turbo-0701). | cs.CL | Python is Not Always the Best Choice: Embracing Multilingual Program of Thoughts |
2024-02-16T13:46:38Z | http://arxiv.org/pdf/2402.10689v2 | http://arxiv.org/abs/2402.10689v2 | Tuan-Phong Nguyen, Simon Razniewski, Gerhard Weikum | 2024-04-17T14:12:07Z | cs.CL | Despite recent progress, large language models (LLMs) still face the
challenge of appropriately reacting to the intricacies of social and cultural
conventions. This paper presents MANGO, a methodology for distilling
high-accuracy, high-recall assertions of cultural knowledge. We judiciously and
iteratively prompt LLMs for this purpose from two entry points, concepts and
cultures. Outputs are consolidated via clustering and generative summarization.
Running the MANGO method with GPT-3.5 as underlying LLM yields 167K
high-accuracy assertions for 30K concepts and 11K cultures, surpassing prior
resources by a large margin. For extrinsic evaluation, we explore augmenting
dialogue systems with cultural knowledge assertions. We find that adding
knowledge from MANGO improves the overall quality, specificity, and cultural
sensitivity of dialogue responses, as judged by human annotators. Data and code
are available for download. | cs.CL | Multi-Cultural Commonsense Knowledge Distillation |
2024-02-16T13:46:06Z | http://arxiv.org/pdf/2402.10688v2 | http://arxiv.org/abs/2402.10688v2 | Haiyan Zhao, Fan Yang, Bo Shen, Himabindu Lakkaraju, Mengnan Du | 2024-04-15T18:43:12Z | cs.CL | Large language models (LLMs) have led to breakthroughs in language tasks, yet
the internal mechanisms that enable their remarkable generalization and
reasoning abilities remain opaque. This lack of transparency presents
challenges such as hallucinations, toxicity, and misalignment with human
values, hindering the safe and beneficial deployment of LLMs. This paper aims
to uncover the mechanisms underlying LLM functionality through the lens of
explainability. First, we review how knowledge is architecturally composed
within LLMs and encoded in their internal parameters via mechanistic
interpretability techniques. Then, we summarize how knowledge is embedded in
LLM representations by leveraging probing techniques and representation
engineering. Additionally, we investigate the training dynamics through a
mechanistic perspective to explain phenomena such as grokking and memorization.
Lastly, we explore how the insights gained from these explanations can enhance
LLM performance through model editing, improve efficiency through pruning, and
better align with human values. | cs.CL | Towards Uncovering How Large Language Model Works: An Explainability Perspective |
2024-02-16T13:39:34Z | http://arxiv.org/pdf/2402.10685v2 | http://arxiv.org/abs/2402.10685v2 | Yi Lu, Xin Zhou, Wei He, Jun Zhao, Tao Ji, Tao Gui, Qi Zhang, Xuanjing Huang | 2024-03-25T11:50:32Z | cs.CL, cs.AI | Large language models (LLMs) have achieved impressive performance in numerous
domains but often struggle to process lengthy inputs effectively and
efficiently due to limited length generalization and attention's quadratic
computational demands. Many sought to mitigate this by restricting the
attention window within the pre-trained length. However, these methods
introduce new issues such as ignoring the middle context and requiring
additional training. To address these problems, we propose LongHeads, a
training-free framework that enhances LLM's long context ability by unlocking
multi-head attention's untapped potential. Instead of allowing each head to
attend to the full sentence, which struggles with generalizing to longer
sequences due to out-of-distribution (OOD) issues, we allow each head to
process in-distribution length by selecting and attending to important context
chunks. To this end, we propose a chunk selection strategy that relies on the
inherent correlation between the query and the key representations, efficiently
distributing context chunks to different heads. In this way, each head ensures
it can effectively process attended tokens within the trained length, while
different heads in different layers can collectively process longer contexts.
LongHeads works efficiently in linear time, fits seamlessly with many LLMs that
use relative positional encoding. LongHeads achieves 100% accuracy at the 128k
length on passkey retrieval task, verifying LongHeads's efficacy in extending
the usable context window for existing models. We release our code at
https://github.com/LuLuLuyi/LongHeads . | cs.CL | LongHeads: Multi-Head Attention is Secretly a Long Context Processor |
2024-02-16T13:28:44Z | http://arxiv.org/pdf/2402.10675v1 | http://arxiv.org/abs/2402.10675v1 | Lars Klöser, Mika Beele, Jan-Niklas Schagen, Bodo Kraft | 2024-02-16T13:28:44Z | cs.CL, I.2.7 | This study pioneers the use of synthetically generated data for training
generative models in document-level text simplification of German texts. We
demonstrate the effectiveness of our approach with real-world online texts.
Addressing the challenge of data scarcity in language simplification, we
crawled professionally simplified German texts and synthesized a corpus using
GPT-4. We finetune Large Language Models with up to 13 billion parameters on
this data and evaluate their performance. This paper employs various
methodologies for evaluation and demonstrates the limitations of currently used
rule-based metrics. Both automatic and manual evaluations reveal that our
models can significantly simplify real-world online texts, indicating the
potential of synthetic data in improving text simplification. | cs.CL | German Text Simplification: Finetuning Large Language Models with Semi-Synthetic Data |
2024-02-16T13:24:05Z | http://arxiv.org/pdf/2402.10671v3 | http://arxiv.org/abs/2402.10671v3 | Yuanzhen Xie, Xinzhou Jin, Tao Xie, MingXiong Lin, Liang Chen, Chenyun Yu, Lei Cheng, ChengXiang Zhuo, Bo Hu, Zang Li | 2024-07-03T04:45:29Z | cs.CL | In-context learning of large-language models (LLMs) has achieved remarkable
success in the field of natural language processing, while extensive case
studies reveal that the single-step chain-of-thought prompting approach faces
challenges such as attention diffusion and inadequate performance in complex
tasks like text-to-SQL. To improve the contextual learning capabilities of LLMs
in text-to-SQL, a workflow paradigm method is proposed, aiming to enhance the
attention and problem-solving scope of LLMs through decomposition.
Specifically, the information determination module for eliminating redundant
information and the brand-new prompt structure based on problem classification
greatly enhance the model's attention. Additionally, the inclusion of
self-correction and active learning modules greatly expands the problem-solving
scope of LLMs, hence improving the upper limit of LLM-based approaches.
Extensive experiments conducted on three datasets demonstrate that our approach
outperforms other methods by a significant margin. About 2-3 percentage point
improvements compared to the existing baseline on the Spider Dev,
Spider-Realistic, and Bird Dev datasets and new SOTA results on the Spider Test
dataset are achieved. Our code is available on GitHub:
\url{https://github.com/FlyingFeather/DEA-SQL}. | cs.CL | Decomposition for Enhancing Attention: Improving LLM-based Text-to-SQL through Workflow Paradigm |
2024-02-16T13:21:33Z | http://arxiv.org/pdf/2402.10670v2 | http://arxiv.org/abs/2402.10670v2 | Yuxuan Kuang, Hai Lin, Meng Jiang | 2024-03-25T02:52:43Z | cs.CL, cs.RO | Object navigation (ObjectNav) requires an agent to navigate through unseen
environments to find queried objects. Many previous methods attempted to solve
this task by relying on supervised or reinforcement learning, where they are
trained on limited household datasets with close-set objects. However, two key
challenges are unsolved: understanding free-form natural language instructions
that demand open-set objects, and generalizing to new environments in a
zero-shot manner. Aiming to solve the two challenges, in this paper, we propose
OpenFMNav, an Open-set Foundation Model based framework for zero-shot object
Navigation. We first unleash the reasoning abilities of large language models
(LLMs) to extract proposed objects from natural language instructions that meet
the user's demand. We then leverage the generalizability of large vision
language models (VLMs) to actively discover and detect candidate objects from
the scene, building a Versatile Semantic Score Map (VSSM). Then, by conducting
common sense reasoning on VSSM, our method can perform effective
language-guided exploration and exploitation of the scene and finally reach the
goal. By leveraging the reasoning and generalizing abilities of foundation
models, our method can understand free-form human instructions and perform
effective open-set zero-shot navigation in diverse environments. Extensive
experiments on the HM3D ObjectNav benchmark show that our method surpasses all
the strong baselines on all metrics, proving our method's effectiveness.
Furthermore, we perform real robot demonstrations to validate our method's
open-set-ness and generalizability to real-world environments. | cs.CL | OpenFMNav: Towards Open-Set Zero-Shot Object Navigation via Vision-Language Foundation Models |
2024-02-16T13:21:06Z | http://arxiv.org/pdf/2402.10669v4 | http://arxiv.org/abs/2402.10669v4 | Guiming Hardy Chen, Shunian Chen, Ziche Liu, Feng Jiang, Benyou Wang | 2024-06-16T12:30:34Z | cs.CL | Adopting human and large language models (LLM) as judges (a.k.a human- and
LLM-as-a-judge) for evaluating the performance of LLMs has recently gained
attention. Nonetheless, this approach concurrently introduces potential biases
from human and LLMs, questioning the reliability of the evaluation results. In
this paper, we propose a novel framework that is free from referencing
groundtruth annotations for investigating Misinformation Oversight Bias, Gender
Bias, Authority Bias and Beauty Bias on LLM and human judges. We curate a
dataset referring to the revised Bloom's Taxonomy and conduct thousands of
evaluations. Results show that human and LLM judges are vulnerable to
perturbations to various degrees, and that even the cutting-edge judges possess
considerable biases. We further exploit these biases to conduct attacks on LLM
judges. We hope that our work can notify the community of the bias and
vulnerability of human- and LLM-as-a-judge, as well as the urgency of
developing robust evaluation systems. | cs.CL | Humans or LLMs as the Judge? A Study on Judgement Biases |
2024-02-16T13:14:35Z | http://arxiv.org/pdf/2402.10666v2 | http://arxiv.org/abs/2402.10666v2 | Xuanliang Zhang, Dingzirui Wang, Longxu Dou, Qingfu Zhu, Wanxiang Che | 2024-06-19T15:19:16Z | cs.CL | Open-domain text-to-SQL is an important task that retrieves question-relevant
tables from massive databases and then generates SQL. However, existing
retrieval methods that retrieve in a single hop do not pay attention to the
text-to-SQL challenge of schema linking, which is aligning the entities in the
question with table entities, reflected in two aspects: similar irrelevant
entity and domain mismatch entity. Therefore, we propose our method, the
multi-hop table retrieval with rewrite and beam search (Murre). To reduce the
effect of the similar irrelevant entity, our method focuses on unretrieved
entities at each hop and considers the low-ranked tables by beam search. To
alleviate the limitation of domain mismatch entity, Murre rewrites the question
based on retrieved tables in multiple hops, decreasing the domain gap with
relevant tables. We conduct experiments on SpiderUnion and BirdUnion+, reaching
new state-of-the-art results with an average improvement of 6.38%. | cs.CL | Multi-Hop Table Retrieval for Open-Domain Text-to-SQL |
2024-02-16T13:13:18Z | http://arxiv.org/pdf/2402.10663v3 | http://arxiv.org/abs/2402.10663v3 | Dingzirui Wang, Longxu Dou, Xuanliang Zhang, Qingfu Zhu, Wanxiang Che | 2024-06-26T06:54:35Z | cs.CL | Currently, the in-context learning method based on large language models
(LLMs) has become the mainstream of text-to-SQL research. Previous works have
discussed how to select demonstrations related to the user question from a
human-labeled demonstration pool. However, human labeling suffers from the
limitations of insufficient diversity and high labeling overhead. Therefore, in
this paper, we discuss how to measure and improve the diversity of the
demonstrations for text-to-SQL. We present a metric to measure the diversity of
the demonstrations and analyze the insufficient of the existing labeled data by
experiments. Based on the above discovery, we propose fusing iteratively for
demonstrations (Fused) to build a high-diversity demonstration pool through
human-free multiple-iteration synthesis, improving diversity and lowering label
cost. Our method achieves an average improvement of 3.2% and 5.0% with and
without human labeling on several mainstream datasets, which proves the
effectiveness of Fused. | cs.CL | Improving Demonstration Diversity by Human-Free Fusing for Text-to-SQL |
2024-02-16T13:11:13Z | http://arxiv.org/pdf/2402.10662v1 | http://arxiv.org/abs/2402.10662v1 | Aravinth Sivaganeshan, Nisansa de Silva | 2024-02-16T13:11:13Z | cs.CL | Named Entity Recognition (NER) is a sequence classification Natural Language
Processing task where entities are identified in the text and classified into
predefined categories. It acts as a foundation for most information extraction
systems. Dungeons and Dragons (D&D) is an open-ended tabletop fantasy game with
its own diverse lore. DnD entities are domain-specific and are thus
unrecognizable by even the state-of-the-art off-the-shelf NER systems as the
NER systems are trained on general data for pre-defined categories such as:
person (PERS), location (LOC), organization (ORG), and miscellaneous (MISC).
For meaningful extraction of information from fantasy text, the entities need
to be classified into domain-specific entity categories as well as the models
be fine-tuned on a domain-relevant corpus. This work uses available lore of
monsters in the D&D domain to fine-tune Trankit, which is a prolific NER
framework that uses a pre-trained model for NER. Upon this training, the system
acquires the ability to extract monster names from relevant domain documents
under a novel NER tag. This work compares the accuracy of the monster name
identification against; the zero-shot Trankit model and two FLAIR models. The
fine-tuned Trankit model achieves an 87.86% F1 score surpassing all the other
considered models. | cs.CL | Fine Tuning Named Entity Extraction Models for the Fantasy Domain |
2024-02-16T13:10:14Z | http://arxiv.org/pdf/2402.10659v3 | http://arxiv.org/abs/2402.10659v3 | Marios Papachristou, Yuan Yuan | 2024-06-02T13:50:14Z | cs.SI, cs.AI, cs.CL, cs.MA | Social networks shape opinions, behaviors, and information dissemination in
human societies. As large language models (LLMs) increasingly integrate into
social and professional environments, understanding their behavior within the
context of social interactions and networks becomes essential. Our study
analyzes LLMs' network formation behavior to examine whether the dynamics of
multiple LLMs are similar to or different from human social dynamics. We
observe that LLMs exhibit key social network principles, including preferential
attachment, triadic closure, homophily, community structure, and the
small-world phenomenon, when asked about their preferences in network
formation. We also investigate LLMs' decision-making based on real-world
networks, revealing that triadic closure and homophily have a stronger
influence than preferential attachment and that LLMs perform well in network
formation predictions. Overall, our study opens up new possibilities for using
LLMs in network science research and helps develop socially aware LLMs by
shedding light on their social interaction behaviors and exploring their
impacts on social dynamics. | cs.SI | Network Formation and Dynamics Among Multi-LLMs |
2024-02-16T13:02:11Z | http://arxiv.org/pdf/2402.10654v1 | http://arxiv.org/abs/2402.10654v1 | Dingzirui Wang, Longxu Dou, Xuanliang Zhang, Qingfu Zhu, Wanxiang Che | 2024-02-16T13:02:11Z | cs.CL | Numerical reasoning is an essential ability for NLP systems to handle numeric
information. Recent research indicates that fine-tuning a small-scale model to
learn generating reasoning processes alongside answers can significantly
enhance performance. However, current methods have the limitation that most
methods generate reasoning processes with large language models (LLMs), which
are "unreliable" since such processes could contain information unrelated to
the answer. To address this limitation, we introduce Enhancing NumeriCal
reasOning with Reliable procEsses (Encore), which derives the reliable
reasoning process by decomposing the answer formula, ensuring which fully
supports the answer. Nevertheless, models could lack enough data to learn the
reasoning process generation adequately, since our method generates only one
single reasoning process for one formula. To overcome this difficulty, we
present a series of pre-training tasks to help models learn the reasoning
process generation with synthesized data. The experiments show that Encore
yields improvement on all five experimental datasets with an average of 1.8%,
proving the effectiveness of our method. | cs.CL | Enhancing Numerical Reasoning with the Guidance of Reliable Reasoning Processes |
2024-02-16T12:47:11Z | http://arxiv.org/pdf/2402.10646v2 | http://arxiv.org/abs/2402.10646v2 | Zhaowei Wang, Wei Fan, Qing Zong, Hongming Zhang, Sehyun Choi, Tianqing Fang, Xin Liu, Yangqiu Song, Ginny Y. Wong, Simon See | 2024-06-17T16:39:02Z | cs.CL | Abstraction ability is crucial in human intelligence, which can also benefit
various tasks in NLP study. Existing work shows that LLMs are deficient in
abstract ability, and how to improve it remains unexplored. In this work, we
design the framework AbsInstruct to enhance LLMs' abstraction ability through
instruction tuning. The framework builds instructions with in-depth
explanations to assist LLMs in capturing the underlying rationale of
abstraction. Meanwhile, we introduce a plausibility estimator to select
instructions that are more consistent with the abstraction knowledge of LLMs to
be aligned. Then, our framework combines abstraction instructions with
general-purpose ones to build a hybrid dataset. Extensive experiments and
analyses demonstrate that our framework can considerably enhance LLMs'
abstraction ability with strong generalization performance while maintaining
their general instruction-following abilities. | cs.CL | AbsInstruct: Eliciting Abstraction Ability from LLMs through Explanation Tuning with Plausibility Estimation |
2024-02-16T12:46:16Z | http://arxiv.org/pdf/2402.10645v1 | http://arxiv.org/abs/2402.10645v1 | Yoonjeong Park, Hyunjin Kim, Chanyeol Choi, Junseong Kim, Jy-yong Sohn | 2024-02-16T12:46:16Z | cs.CL, cs.AI | Chain-of-thought (CoT) prompting is a simple and effective method for
improving the reasoning capabilities of Large language models (LLMs). The basic
idea of CoT is to let LLMs break down their thought processes step-by-step by
putting exemplars in the input prompt. However, the densely structured prompt
exemplars of CoT may cause the cognitive overload of LLMs. Inspired by human
cognition, we introduce CoT-Sep, a novel method that strategically employs
separators at the end of each exemplar in CoT prompting. These separators are
designed to help the LLMs understand their thought processes better while
reasoning. It turns out that CoT-Sep significantly improves the LLMs'
performances on complex reasoning tasks (e.g., GSM-8K, AQuA, CSQA), compared
with the vanilla CoT, which does not use separators. We also study the effects
of the type and the location of separators tested on multiple LLMs, including
GPT-3.5-Turbo, GPT-4, and LLaMA-2 7B. Interestingly, the type/location of
separators should be chosen appropriately to boost the reasoning capability of
CoT. | cs.CL | Can Separators Improve Chain-of-Thought Prompting? |
2024-02-16T12:44:15Z | http://arxiv.org/pdf/2402.10644v2 | http://arxiv.org/abs/2402.10644v2 | Yaroslav Aksenov, Nikita Balagansky, Sofia Maria Lo Cicero Vaina, Boris Shaposhnikov, Alexey Gorbatovski, Daniil Gavrilov | 2024-06-05T14:13:22Z | cs.LG, cs.CL | Advancing the frontier of subquadratic architectures for Language Models
(LMs) is crucial in the rapidly evolving field of natural language processing.
Current innovations, including State Space Models, were initially celebrated
for surpassing Transformer performance on language modeling tasks. However,
these models have revealed deficiencies in essential In-Context Learning
capabilities - a domain where the Transformer traditionally shines. The Based
model emerged as a hybrid solution, blending a Linear Transformer with a kernel
inspired by the Taylor expansion of exponential functions, augmented by
convolutional networks. Mirroring the Transformer's in-context adeptness, it
became a strong contender in the field. In our work, we present a singular,
elegant alteration to the Based kernel that amplifies its In-Context Learning
abilities evaluated with the Multi-Query Associative Recall task and overall
language modeling process, as demonstrated on the Pile dataset. | cs.LG | Linear Transformers with Learnable Kernel Functions are Better In-Context Models |
2024-02-16T12:43:26Z | http://arxiv.org/pdf/2402.10643v1 | http://arxiv.org/abs/2402.10643v1 | Ronald Cardenas, Matthias Galle, Shay B. Cohen | 2024-02-16T12:43:26Z | cs.CL, cs.AI | Extractive summaries are usually presented as lists of sentences with no
expected cohesion between them. In this paper, we aim to enforce cohesion
whilst controlling for informativeness and redundancy in summaries, in cases
where the input exhibits high redundancy. The pipeline controls for redundancy
in long inputs as it is consumed, and balances informativeness and cohesion
during sentence selection. Our sentence selector simulates human memory to keep
track of topics --modeled as lexical chains--, enforcing cohesive ties between
noun phrases. Across a variety of domains, our experiments revealed that it is
possible to extract highly cohesive summaries that nevertheless read as
informative to humans as summaries extracted by only accounting for
informativeness or redundancy. The extracted summaries exhibit smooth topic
transitions between sentences as signaled by lexical chains, with chains
spanning adjacent or near-adjacent sentences. | cs.CL | `Keep it Together': Enforcing Cohesion in Extractive Summaries by Simulating Human Memory |
2024-02-16T12:39:10Z | http://arxiv.org/pdf/2402.10639v2 | http://arxiv.org/abs/2402.10639v2 | Tuc Nguyen, Thai Le | 2024-06-06T04:52:09Z | cs.CL | Several parameter-efficient fine-tuning methods based on adapters have been
proposed as a streamlined approach to incorporate not only a single specialized
knowledge into existing Pre-Trained Language Models (PLMs) but also multiple of
them at once. Recent works such as AdapterSoup propose to mix not all but only
a selective sub-set of domain-specific adapters during inference via model
weight averaging to optimize performance on novel, unseen domains with
excellent computational efficiency. However, the essential generalizability of
this emerging weight-space adapter mixing mechanism on \textit{unseen,
in-domain examples} remains unexplored. Thus, in this study, we conduct a
comprehensive analysis to elucidate the generalizability of domain-specific
adapter mixtures in in-domain evaluation. We also provide investigations into
the inner workings of the mixture of domain-specific adapters by analyzing
their weight signs, yielding critical analysis on the negative correlation
between their fraction of weight sign difference and their mixtures'
generalizability. | cs.CL | Generalizability of Mixture of Domain-Specific Adapters from the Lens of Signed Weight Directions and its Application to Effective Model Pruning |
2024-02-16T12:27:15Z | http://arxiv.org/pdf/2402.10631v1 | http://arxiv.org/abs/2402.10631v1 | Dayou Du, Yijia Zhang, Shijie Cao, Jiaqi Guo, Ting Cao, Xiaowen Chu, Ningyi Xu | 2024-02-16T12:27:15Z | cs.CL | The upscaling of Large Language Models (LLMs) has yielded impressive advances
in natural language processing, yet it also poses significant deployment
challenges. Weight quantization has emerged as a widely embraced solution to
reduce memory and computational demands. This paper introduces BitDistiller, a
framework that synergizes Quantization-Aware Training (QAT) with Knowledge
Distillation (KD) to boost the performance of LLMs at ultra-low precisions
(sub-4-bit). Specifically, BitDistiller first incorporates a tailored
asymmetric quantization and clipping technique to maximally preserve the
fidelity of quantized weights, and then proposes a novel Confidence-Aware
Kullback-Leibler Divergence (CAKLD) objective, which is employed in a
self-distillation manner to enable faster convergence and superior model
performance. Empirical evaluations demonstrate that BitDistiller significantly
surpasses existing methods in both 3-bit and 2-bit configurations on general
language understanding and complex reasoning benchmarks. Notably, BitDistiller
is shown to be more cost-effective, demanding fewer data and training
resources. The code is available at https://github.com/DD-DuDa/BitDistiller. | cs.CL | BitDistiller: Unleashing the Potential of Sub-4-Bit LLMs via Self-Distillation |
2024-02-16T12:12:05Z | http://arxiv.org/pdf/2402.10618v2 | http://arxiv.org/abs/2402.10618v2 | Yihong Tang, Jiao Ou, Che Liu, Fuzheng Zhang, Di Zhang, Kun Gai | 2024-06-15T13:51:38Z | cs.CL | The advent of Large Language Models (LLMs) has propelled dialogue generation
into new realms, particularly in the field of role-playing systems (RPSs).
While enhanced with ordinary role-relevant training dialogues, existing
LLM-based RPSs still struggle to align with roles when handling intricate and
trapped queries in boundary scenarios. In this paper, we design the Modular
ORchestrated Trap-setting Interaction SystEm (MORTISE) to benchmark and improve
the role-playing LLMs' performance. MORTISE can produce highly role-relevant
aggressive queries through the collaborative effort of multiple LLM-based
modules, and formulate corresponding responses to create an adversarial
training dataset via a consistent response generator. We select 190 Chinese and
English roles to construct aggressive queries to benchmark existing
role-playing LLMs. Through comprehensive evaluation, we find that existing
models exhibit a general deficiency in role alignment capabilities. We further
select 180 of the roles to collect an adversarial training dataset (named
RoleAD) and retain the other 10 roles for testing. Experiments on models
improved by RoleAD indicate that our adversarial dataset ameliorates this
deficiency, with the improvements demonstrating a degree of generalizability in
ordinary scenarios. | cs.CL | Enhancing Role-playing Systems through Aggressive Queries: Evaluation and Improvement |
2024-02-16T12:00:34Z | http://arxiv.org/pdf/2402.10614v2 | http://arxiv.org/abs/2402.10614v2 | Ming Li, Jiuhai Chen, Lichang Chen, Tianyi Zhou | 2024-06-07T20:19:09Z | cs.CL, cs.AI, cs.LG | Making LLMs speak for different, especially minority groups of people, and
generate statements supporting their diverse or even controversial perspectives
is critical to creating an inclusive environment. However, existing LLMs lack
sufficient controllability to the stance of their generated content, which
often contains inconsistent, neutral, or biased statements. In this paper, we
improve the controllability of LLMs in generating statements supporting an
argument the user defined in the prompt. We find that multi-round debates
between two LLMs with opposite stances generate higher-quality and more salient
statements for each, which are important training data to improve the
controllability of LLMs. Motivated by this, we develop a novel debate & tuning
(DEBATUNE) pipeline finetuning LLMs to generate the statements obtained via
debate. To examine DEBATUNE, we curate the largest dataset of debate topics so
far, which covers 710 controversial topics and corresponding arguments for each
topic. Evaluations by the GPT-4 judge with a novel controversy controllability
metric show that LLMs' capability of generating diverse perspectives is
significantly improved by DEBATUNE. Moreover, such controllability can be
generalized to unseen topics, generating high-quality statements supporting
controversial arguments. | cs.CL | Can LLMs Speak For Diverse People? Tuning LLMs via Debate to Generate Controllable Controversial Statements |
2024-02-16T11:55:40Z | http://arxiv.org/pdf/2402.10612v1 | http://arxiv.org/abs/2402.10612v1 | Hanxing Ding, Liang Pang, Zihao Wei, Huawei Shen, Xueqi Cheng | 2024-02-16T11:55:40Z | cs.CL | Hallucinations pose a significant challenge for the practical implementation
of large language models (LLMs). The utilization of parametric knowledge in
generating factual content is constrained by the limited knowledge of LLMs,
potentially resulting in internal hallucinations. While incorporating external
information can help fill knowledge gaps, it also introduces the risk of
irrelevant information, thereby increasing the likelihood of external
hallucinations. A careful and balanced integration of the parametric knowledge
within LLMs with external information is crucial to alleviate hallucinations.
In this study, we present Rowen, a novel approach that enhances LLMs with a
selective retrieval augmentation process tailored to address hallucinated
outputs. This process is governed by a multilingual semantic-aware detection
module, which evaluates the consistency of the perturbed responses across
various languages for the same queries. Upon detecting inconsistencies
indicative of hallucinations, Rowen activates the retrieval of external
information to rectify the model outputs. Rowen adeptly harmonizes the
intrinsic parameters in LLMs with external knowledge sources, effectively
mitigating hallucinations by ensuring a balanced integration of internal
reasoning and external evidence. Through a comprehensive empirical analysis, we
demonstrate that Rowen surpasses the current state-of-the-art in both detecting
and mitigating hallucinated content within the outputs of LLMs. | cs.CL | Retrieve Only When It Needs: Adaptive Retrieval Augmentation for Hallucination Mitigation in Large Language Models |
2024-02-16T11:37:05Z | http://arxiv.org/pdf/2402.10601v1 | http://arxiv.org/abs/2402.10601v1 | Divij Handa, Advait Chirmule, Bimal Gajera, Chitta Baral | 2024-02-16T11:37:05Z | cs.CL, cs.AI | Large Language Models (LLMs) are aligned to moral and ethical guidelines but
remain susceptible to creative prompts called Jailbreak that can bypass the
alignment process. However, most jailbreaking prompts contain harmful questions
in the natural language (mainly English), which can be detected by the LLM
themselves. In this paper, we present jailbreaking prompts encoded using
cryptographic techniques. We first present a pilot study on the
state-of-the-art LLM, GPT-4, in decoding several safe sentences that have been
encrypted using various cryptographic techniques and find that a
straightforward word substitution cipher can be decoded most effectively.
Motivated by this result, we use this encoding technique for writing
jailbreaking prompts. We present a mapping of unsafe words with safe words and
ask the unsafe question using these mapped words. Experimental results show an
attack success rate (up to 59.42%) of our proposed jailbreaking approach on
state-of-the-art proprietary models including ChatGPT, GPT-4, and Gemini-Pro.
Additionally, we discuss the over-defensiveness of these models. We believe
that our work will encourage further research in making these LLMs more robust
while maintaining their decoding capabilities. | cs.CL | Jailbreaking Proprietary Large Language Models using Word Substitution Cipher |
2024-02-16T11:30:11Z | http://arxiv.org/pdf/2402.10597v1 | http://arxiv.org/abs/2402.10597v1 | Niall Taylor, Upamanyu Ghose, Omid Rohanian, Mohammadmahdi Nouriborji, Andrey Kormilitzin, David Clifton, Alejo Nevado-Holgado | 2024-02-16T11:30:11Z | cs.CL, cs.AI | The entry of large language models (LLMs) into research and commercial spaces
has led to a trend of ever-larger models, with initial promises of
generalisability, followed by a widespread desire to downsize and create
specialised models without the need for complete fine-tuning, using Parameter
Efficient Fine-tuning (PEFT) methods. We present an investigation into the
suitability of different PEFT methods to clinical decision-making tasks, across
a range of model sizes, including extremely small models with as few as $25$
million parameters.
Our analysis shows that the performance of most PEFT approaches varies
significantly from one task to another, with the exception of LoRA, which
maintains relatively high performance across all model sizes and tasks,
typically approaching or matching full fine-tuned performance. The
effectiveness of PEFT methods in the clinical domain is evident, particularly
for specialised models which can operate on low-cost, in-house computing
infrastructure. The advantages of these models, in terms of speed and reduced
training costs, dramatically outweighs any performance gain from large
foundation LLMs. Furthermore, we highlight how domain-specific pre-training
interacts with PEFT methods and model size, and discuss how these factors
interplay to provide the best efficiency-performance trade-off. Full code
available at: tbd. | cs.CL | Efficiency at Scale: Investigating the Performance of Diminutive Language Models in Clinical Tasks |
2024-02-16T11:21:28Z | http://arxiv.org/pdf/2402.10588v4 | http://arxiv.org/abs/2402.10588v4 | Chris Wendler, Veniamin Veselovsky, Giovanni Monea, Robert West | 2024-06-08T11:15:14Z | cs.CL, cs.CY | We ask whether multilingual language models trained on unbalanced,
English-dominated corpora use English as an internal pivot language -- a
question of key importance for understanding how language models function and
the origins of linguistic bias. Focusing on the Llama-2 family of transformer
models, our study uses carefully constructed non-English prompts with a unique
correct single-token continuation. From layer to layer, transformers gradually
map an input embedding of the final prompt token to an output embedding from
which next-token probabilities are computed. Tracking intermediate embeddings
through their high-dimensional space reveals three distinct phases, whereby
intermediate embeddings (1) start far away from output token embeddings; (2)
already allow for decoding a semantically correct next token in the middle
layers, but give higher probability to its version in English than in the input
language; (3) finally move into an input-language-specific region of the
embedding space. We cast these results into a conceptual model where the three
phases operate in "input space", "concept space", and "output space",
respectively. Crucially, our evidence suggests that the abstract "concept
space" lies closer to English than to other languages, which may have important
consequences regarding the biases held by multilingual language models. | cs.CL | Do Llamas Work in English? On the Latent Language of Multilingual Transformers |
2024-02-16T11:20:30Z | http://arxiv.org/pdf/2402.10586v2 | http://arxiv.org/abs/2402.10586v2 | Zae Myung Kim, Kwang Hee Lee, Preston Zhu, Vipul Raheja, Dongyeop Kang | 2024-06-06T20:04:52Z | cs.CL, cs.AI | With the advent of large language models (LLM), the line between
human-crafted and machine-generated texts has become increasingly blurred. This
paper delves into the inquiry of identifying discernible and unique linguistic
properties in texts that were written by humans, particularly uncovering the
underlying discourse structures of texts beyond their surface structures.
Introducing a novel methodology, we leverage hierarchical parse trees and
recursive hypergraphs to unveil distinctive discourse patterns in texts
produced by both LLMs and humans. Empirical findings demonstrate that, although
both LLMs and humans generate distinct discourse patterns influenced by
specific domains, human-written texts exhibit more structural variability,
reflecting the nuanced nature of human writing in different domains. Notably,
incorporating hierarchical discourse features enhances binary classifiers'
overall performance in distinguishing between human-written and
machine-generated texts, even on out-of-distribution and paraphrased samples.
This underscores the significance of incorporating hierarchical discourse
features in the analysis of text patterns. The code and dataset are available
at https://github.com/minnesotanlp/threads-of-subtlety. | cs.CL | Threads of Subtlety: Detecting Machine-Generated Texts Through Discourse Motifs |
2024-02-16T11:02:29Z | http://arxiv.org/pdf/2402.10573v2 | http://arxiv.org/abs/2402.10573v2 | Zhen Zhang, Yuhua Zhao, Hang Gao, Mengting Hu | 2024-02-27T15:08:02Z | cs.CL, I.2.7 | Named Entity Recognition (NER) serves as a fundamental task in natural
language understanding, bearing direct implications for web content analysis,
search engines, and information retrieval systems. Fine-tuned NER models
exhibit satisfactory performance on standard NER benchmarks. However, due to
limited fine-tuning data and lack of knowledge, it performs poorly on unseen
entity recognition. As a result, the usability and reliability of NER models in
web-related applications are compromised. Instead, Large Language Models (LLMs)
like GPT-4 possess extensive external knowledge, but research indicates that
they lack specialty for NER tasks. Furthermore, non-public and large-scale
weights make tuning LLMs difficult. To address these challenges, we propose a
framework that combines small fine-tuned models with LLMs (LinkNER) and an
uncertainty-based linking strategy called RDC that enables fine-tuned models to
complement black-box LLMs, achieving better performance. We experiment with
both standard NER test sets and noisy social media datasets. LinkNER enhances
NER task performance, notably surpassing SOTA models in robustness tests. We
also quantitatively analyze the influence of key components like uncertainty
estimation methods, LLMs, and in-context learning on diverse NER tasks,
offering specific web-related recommendations. | cs.CL | LinkNER: Linking Local Named Entity Recognition Models to Large Language Models using Uncertainty |
2024-02-16T10:55:38Z | http://arxiv.org/pdf/2402.10571v2 | http://arxiv.org/abs/2402.10571v2 | Afra Amini, Tim Vieira, Ryan Cotterell | 2024-06-06T12:02:37Z | cs.CL, cs.AI, cs.LG | Direct preference optimization (DPO) is a successful fine-tuning strategy for
aligning large language models with human preferences without the need to train
a reward model or employ reinforcement learning. DPO, as originally formulated,
relies on binary preference data and fine-tunes a language model to increase
the likelihood of a preferred response over a dispreferred response. However,
not all preference pairs are equal. Sometimes, the preferred response is only
slightly better than the dispreferred one. In other cases, the preference is
much stronger. For instance, if a response contains harmful or toxic content,
the annotator will have a strong preference for that response. In this paper,
we propose a generalization of DPO, termed DPO with an offset (ODPO), that does
not treat every preference pair equally during fine-tuning. Intuitively, ODPO
requires the difference between the likelihood of the preferred and
dispreferred response to be greater than an offset value. The offset is
determined based on the extent to which one response is preferred over another.
Our experiments on various tasks suggest that ODPO significantly outperforms
DPO in aligning language models, especially when the number of preference pairs
is limited. | cs.CL | Direct Preference Optimization with an Offset |
2024-02-16T10:54:10Z | http://arxiv.org/pdf/2402.10567v4 | http://arxiv.org/abs/2402.10567v4 | Yogesh Tripathi, Raghav Donakanti, Sahil Girhepuje, Ishan Kavathekar, Bhaskara Hanuma Vedula, Gokul S Krishnan, Shreya Goyal, Anmol Goel, Balaraman Ravindran, Ponnurangam Kumaraguru | 2024-06-17T17:46:07Z | cs.CL, cs.AI | Recent advancements in language technology and Artificial Intelligence have
resulted in numerous Language Models being proposed to perform various tasks in
the legal domain ranging from predicting judgments to generating summaries.
Despite their immense potential, these models have been proven to learn and
exhibit societal biases and make unfair predictions. In this study, we explore
the ability of Large Language Models (LLMs) to perform legal tasks in the
Indian landscape when social factors are involved. We present a novel metric,
$\beta$-weighted $\textit{Legal Safety Score ($LSS_{\beta}$)}$, which
encapsulates both the fairness and accuracy aspects of the LLM. We assess LLMs'
safety by considering its performance in the $\textit{Binary Statutory
Reasoning}$ task and its fairness exhibition with respect to various axes of
disparities in the Indian society. Task performance and fairness scores of
LLaMA and LLaMA--2 models indicate that the proposed $LSS_{\beta}$ metric can
effectively determine the readiness of a model for safe usage in the legal
sector. We also propose finetuning pipelines, utilising specialised legal
datasets, as a potential method to mitigate bias and improve model safety. The
finetuning procedures on LLaMA and LLaMA--2 models increase the $LSS_{\beta}$,
improving their usability in the Indian legal domain. Our code is publicly
released. | cs.CL | InSaAF: Incorporating Safety through Accuracy and Fairness | Are LLMs ready for the Indian Legal Domain? |
2024-02-16T10:40:38Z | http://arxiv.org/pdf/2402.10558v1 | http://arxiv.org/abs/2402.10558v1 | Achille Globo, Antonio Trevisi, Andrea Zugarini, Leonardo Rigutini, Marco Maggini, Stefano Melacci | 2024-02-16T10:40:38Z | cs.CL | Paraphrasing is the task of re-writing an input text using other words,
without altering the meaning of the original content. Conversational systems
can exploit automatic paraphrasing to make the conversation more natural, e.g.,
talking about a certain topic using different paraphrases in different time
instants. Recently, the task of automatically generating paraphrases has been
approached in the context of Natural Language Generation (NLG). While many
existing systems simply consist in rule-based models, the recent success of the
Deep Neural Networks in several NLG tasks naturally suggests the possibility of
exploiting such networks for generating paraphrases. However, the main obstacle
toward neural-network-based paraphrasing is the lack of large datasets with
aligned pairs of sentences and paraphrases, that are needed to efficiently
train the neural models. In this paper we present a method for the automatic
generation of large aligned corpora, that is based on the assumption that news
and blog websites talk about the same events using different narrative styles.
We propose a similarity search procedure with linguistic constraints that,
given a reference sentence, is able to locate the most similar candidate
paraphrases out from millions of indexed sentences. The data generation process
is evaluated in the case of the Italian language, performing experiments using
pointer-based deep neural architectures. | cs.CL | Neural paraphrasing by automatically crawled and aligned sentence pairs |
2024-02-16T10:36:38Z | http://arxiv.org/pdf/2402.10555v2 | http://arxiv.org/abs/2402.10555v2 | Chiyu Zhang, Yifei Sun, Jun Chen, Jie Lei, Muhammad Abdul-Mageed, Sinong Wang, Rong Jin, Sem Park, Ning Yao, Bo Long | 2024-05-22T02:44:40Z | cs.IR, cs.CL | Leveraging users' long engagement histories is essential for personalized
content recommendations. The success of pretrained language models (PLMs) in
NLP has led to their use in encoding user histories and candidate items,
framing content recommendations as textual semantic matching tasks. However,
existing works still struggle with processing very long user historical text
and insufficient user-item interaction. In this paper, we introduce a
content-based recommendation framework, SPAR, which effectively tackles the
challenges of holistic user interest extraction from the long user engagement
history. It achieves so by leveraging PLM, poly-attention layers and attention
sparsity mechanisms to encode user's history in a session-based manner. The
user and item side features are sufficiently fused for engagement prediction
while maintaining standalone representations for both sides, which is efficient
for practical model deployment. Moreover, we enhance user profiling by
exploiting large language model (LLM) to extract global interests from user
engagement history. Extensive experiments on two benchmark datasets demonstrate
that our framework outperforms existing state-of-the-art (SoTA) methods. | cs.IR | SPAR: Personalized Content-Based Recommendation via Long Engagement Attention |
2024-02-16T10:35:18Z | http://arxiv.org/pdf/2402.10554v2 | http://arxiv.org/abs/2402.10554v2 | Xiaobo Guo, Soroush Vosoughi | 2024-06-17T20:06:04Z | cs.CL | Aspect-based summarization has seen significant advancements, especially in
structured text. Yet, summarizing disordered, large-scale texts, like those
found in social media and customer feedback, remains a significant challenge.
Current research largely targets predefined aspects within structured texts,
neglecting the complexities of dynamic and disordered environments. Addressing
this gap, we introduce Disordered-DABS, a novel benchmark for dynamic
aspect-based summarization tailored to unstructured text. Developed by adapting
existing datasets for cost-efficiency and scalability, our comprehensive
experiments and detailed human evaluations reveal that Disordered-DABS poses
unique challenges to contemporary summarization models, including
state-of-the-art language models such as GPT-3.5. | cs.CL | Disordered-DABS: A Benchmark for Dynamic Aspect-Based Summarization in Disordered Texts |
2024-02-16T10:35:01Z | http://arxiv.org/pdf/2402.10553v1 | http://arxiv.org/abs/2402.10553v1 | Andrea Pazienza, Nicola Macchiarulo, Felice Vitulano, Antonio Fiorentini, Marco Cammisa, Leonardo Rigutini, Ernesto Di Iorio, Achille Globo, Antonio Trevisi | 2024-02-16T10:35:01Z | cs.RO, cs.CL, cs.CV, cs.LG | From robots that replace workers to robots that serve as helpful colleagues,
the field of robotic automation is experiencing a new trend that represents a
huge challenge for component manufacturers. The contribution starts from an
innovative vision that sees an ever closer collaboration between Cobot, able to
do a specific physical job with precision, the AI world, able to analyze
information and support the decision-making process, and the man able to have a
strategic vision of the future. | cs.RO | A novel integrated industrial approach with cobots in the age of industry 4.0 through conversational interaction and computer vision |
2024-02-16T10:32:16Z | http://arxiv.org/pdf/2402.10552v3 | http://arxiv.org/abs/2402.10552v3 | Minghan Wang, Thuy-Trang Vu, Yuxia Wang, Ehsan Shareghi, Gholamreza Haffari | 2024-06-21T07:18:13Z | cs.CL | Simultaneous machine translation (SimulMT) presents a challenging trade-off
between translation quality and latency. Recent studies have shown that LLMs
can achieve good performance in SimulMT tasks. However, this often comes at the
expense of high inference cost and latency. In this paper, we propose a
conversational SimulMT framework to enhance the inference efficiency of
LLM-based SimulMT through multi-turn-dialogue-based decoding. Our experiments
with Llama2-7b-chat on two SimulMT benchmarks demonstrate the superiority of
LLM in translation quality while achieving comparable computational latency to
specialized SimulMT models. | cs.CL | Conversational SimulMT: Efficient Simultaneous Translation with Large Language Models |
2024-02-16T10:11:20Z | http://arxiv.org/pdf/2402.10543v1 | http://arxiv.org/abs/2402.10543v1 | Nicholas Asher, Swarnadeep Bhar | 2024-02-16T10:11:20Z | cs.CL, cs.AI, I.2.7 | Despite great performance on many tasks, language models (LMs) still struggle
with reasoning, sometimes providing responses that cannot possibly be true
because they stem from logical incoherence. We call such responses
\textit{strong hallucinations} and prove that they follow from an LM's
computation of its internal representations for logical operators and outputs
from those representations. Focusing on negation, we provide a novel solution
in which negation is treated not as another element of a latent representation,
but as \textit{an operation over an LM's latent representations that constrains
how they may evolve}. We show that our approach improves model performance in
cloze prompting and natural language inference tasks with negation without
requiring training on sparse negative data. | cs.CL | Strong hallucinations from negation and how to fix them |
2024-02-16T09:37:54Z | http://arxiv.org/pdf/2402.10532v1 | http://arxiv.org/abs/2402.10532v1 | Jenny Kunz, Marco Kuhlmann | 2024-02-16T09:37:54Z | cs.CL, cs.AI, cs.HC, cs.LG | The self-rationalising capabilities of large language models (LLMs) have been
explored in restricted settings, using task/specific data sets. However,
current LLMs do not (only) rely on specifically annotated data; nonetheless,
they frequently explain their outputs. The properties of the generated
explanations are influenced by the pre-training corpus and by the target data
used for instruction fine-tuning. As the pre-training corpus includes a large
amount of human-written explanations "in the wild", we hypothesise that LLMs
adopt common properties of human explanations. By analysing the outputs for a
multi-domain instruction fine-tuning data set, we find that generated
explanations show selectivity and contain illustrative elements, but less
frequently are subjective or misleading. We discuss reasons and consequences of
the properties' presence or absence. In particular, we outline positive and
negative implications depending on the goals and user groups of the
self-rationalising system. | cs.CL | Properties and Challenges of LLM-Generated Explanations |
2024-02-16T09:29:50Z | http://arxiv.org/pdf/2402.10528v2 | http://arxiv.org/abs/2402.10528v2 | Xin Xu, Shizhe Diao, Can Yang, Yang Wang | 2024-06-15T16:53:58Z | cs.CL, cs.AI | Chain-of-Thought (CoT) prompting has marked a significant advancement in
enhancing the reasoning capabilities of large language models (LLMs). Previous
studies have developed various extensions of CoT, which focus primarily on
enhancing end-task performance. In addition, there has been research on
assessing the quality of reasoning chains in CoT. This raises an intriguing
question: Is it possible to predict the accuracy of LLM outputs by scrutinizing
the reasoning chains they generate? To answer this research question, we
introduce a benchmark, R2PE, designed specifically to explore the relationship
between reasoning chains and performance in various reasoning tasks spanning
five different domains. This benchmark aims to measure the falsehood of the
final output of LLMs based on the reasoning steps. To make full use of
information in multiple reasoning chains, we propose the process discernibility
score (PDS) framework that beats the answer-checking baseline by a large
margin. Concretely, this resulted in an average of $5.1\%$ increase in the F1
score and $2.97\%$ improvement in AUC-PR across all 45 subsets within R2PE. We
further demonstrate our PDS's efficacy in advancing open-domain QA accuracy.
Data and code are available at https://github.com/XinXU-USTC/R2PE. | cs.CL | Can We Verify Step by Step for Incorrect Answer Detection? |
2024-02-16T09:29:38Z | http://arxiv.org/pdf/2402.10527v1 | http://arxiv.org/abs/2402.10527v1 | R. Patrick Xian, Alex J. Lee, Vincent Wang, Qiming Cui, Russell Ro, Reza Abbasi-Asl | 2024-02-16T09:29:38Z | cs.CL, cs.CR, stat.AP | The increasing depth of parametric domain knowledge in large language models
(LLMs) is fueling their rapid deployment in real-world applications. In
high-stakes and knowledge-intensive tasks, understanding model vulnerabilities
is essential for quantifying the trustworthiness of model predictions and
regulating their use. The recent discovery of named entities as adversarial
examples in natural language processing tasks raises questions about their
potential guises in other settings. Here, we propose a powerscaled
distance-weighted sampling scheme in embedding space to discover diverse
adversarial entities as distractors. We demonstrate its advantage over random
sampling in adversarial question answering on biomedical topics. Our approach
enables the exploration of different regions on the attack surface, which
reveals two regimes of adversarial entities that markedly differ in their
characteristics. Moreover, we show that the attacks successfully manipulate
token-wise Shapley value explanations, which become deceptive in the
adversarial setting. Our investigations illustrate the brittleness of domain
knowledge in LLMs and reveal a shortcoming of standard evaluations for
high-capacity models. | cs.CL | Zero-shot sampling of adversarial entities in biomedical question answering |
2024-02-16T09:14:49Z | http://arxiv.org/pdf/2402.10524v1 | http://arxiv.org/abs/2402.10524v1 | Minsuk Kahng, Ian Tenney, Mahima Pushkarna, Michael Xieyang Liu, James Wexler, Emily Reif, Krystal Kallarackal, Minsuk Chang, Michael Terry, Lucas Dixon | 2024-02-16T09:14:49Z | cs.HC, cs.AI, cs.CL, cs.LG | Automatic side-by-side evaluation has emerged as a promising approach to
evaluating the quality of responses from large language models (LLMs). However,
analyzing the results from this evaluation approach raises scalability and
interpretability challenges. In this paper, we present LLM Comparator, a novel
visual analytics tool for interactively analyzing results from automatic
side-by-side evaluation. The tool supports interactive workflows for users to
understand when and why a model performs better or worse than a baseline model,
and how the responses from two models are qualitatively different. We
iteratively designed and developed the tool by closely working with researchers
and engineers at a large technology company. This paper details the user
challenges we identified, the design and development of the tool, and an
observational study with participants who regularly evaluate their models. | cs.HC | LLM Comparator: Visual Analytics for Side-by-Side Evaluation of Large Language Models |
2024-02-16T08:19:34Z | http://arxiv.org/pdf/2402.10500v2 | http://arxiv.org/abs/2402.10500v2 | Nirjhar Das, Souradip Chakraborty, Aldo Pacchiano, Sayak Ray Chowdhury | 2024-06-05T15:10:08Z | cs.LG, cs.AI, cs.CL | Reinforcement Learning from Human Feedback (RLHF) is pivotal in aligning
Large Language Models (LLMs) with human preferences. Although aligned
generative models have shown remarkable abilities in various tasks, their
reliance on high-quality human preference data creates a costly bottleneck in
the practical application of RLHF. One primary reason is that current methods
rely on uniformly picking prompt-generation pairs from a dataset of
prompt-generations, to collect human feedback, resulting in sub-optimal
alignment under a constrained budget, which highlights the criticality of
adaptive strategies in efficient alignment. Recent works [Mehta et al., 2023,
Muldrew et al., 2024] have tried to address this problem by designing various
heuristics based on generation uncertainty. However, either the assumptions in
[Mehta et al., 2023] are restrictive, or [Muldrew et al., 2024] do not provide
any rigorous theoretical guarantee. To address these, we reformulate RLHF
within contextual preference bandit framework, treating prompts as contexts,
and develop an active-learning algorithm, $\textit{Active Preference
Optimization}$ ($\texttt{APO}$), which enhances model alignment by querying
preference data from the most important samples, achieving superior performance
for small sample budget. We analyze the theoretical performance guarantees of
$\texttt{APO}$ under the BTL preference model showing that the suboptimality
gap of the policy learned via $\texttt{APO}$ scales as $O(1/\sqrt{T})$ for a
budget of $T$. We also show that collecting preference data by choosing prompts
randomly leads to a policy that suffers a constant sub-optimality. We perform
detailed experimental evaluations on practical preference datasets to validate
$\texttt{APO}$'s efficacy over the existing methods, establishing it as a
sample-efficient and practical solution of alignment in a cost-effective and
scalable manner. | cs.LG | Active Preference Optimization for Sample Efficient RLHF |
2024-02-16T08:10:34Z | http://arxiv.org/pdf/2402.10496v2 | http://arxiv.org/abs/2402.10496v2 | Haoqiang Kang, Terra Blevins, Luke Zettlemoyer | 2024-06-16T00:44:28Z | cs.CL, cs.AI | While many hallucination detection techniques have been evaluated on English
text, their effectiveness in multilingual contexts remains unknown. This paper
assesses how well various factual hallucination detection metrics (lexical
metrics like ROUGE and Named Entity Overlap, and Natural Language Inference
(NLI)-based metrics) identify hallucinations in generated biographical
summaries across languages. We compare how well automatic metrics correlate to
each other and whether they agree with human judgments of factuality. Our
analysis reveals that while the lexical metrics are ineffective, NLI-based
metrics perform well, correlating with human annotations in many settings and
often outperforming supervised models. However, NLI metrics are still limited,
as they do not detect single-fact hallucinations well and fail for
lower-resource languages. Therefore, our findings highlight the gaps in
exisiting hallucination detection methods for non-English languages and
motivate future research to develop more robust multilingual detection methods
for LLM hallucinations. | cs.CL | Comparing Hallucination Detection Metrics for Multilingual Generation |
2024-02-16T07:05:49Z | http://arxiv.org/pdf/2402.10481v2 | http://arxiv.org/abs/2402.10481v2 | Xiaorui Zuo, Yao-Tsung Chen, Wolfgang Karl Härdle | 2024-05-04T14:32:44Z | q-fin.CP, cs.AI, cs.CL, cs.LG, q-fin.ST | In the burgeoning realm of cryptocurrency, social media platforms like
Twitter have become pivotal in influencing market trends and investor
sentiments. In our study, we leverage GPT-4 and a fine-tuned transformer-based
BERT model for a multimodal sentiment analysis, focusing on the impact of emoji
sentiment on cryptocurrency markets. By translating emojis into quantifiable
sentiment data, we correlate these insights with key market indicators like BTC
Price and the VCRIX index. Our architecture's analysis of emoji sentiment
demonstrated a distinct advantage over FinBERT's pure text sentiment analysis
in such predicting power. This approach may be fed into the development of
trading strategies aimed at utilizing social media elements to identify and
forecast market trends. Crucially, our findings suggest that strategies based
on emoji sentiment can facilitate the avoidance of significant market downturns
and contribute to the stabilization of returns. This research underscores the
practical benefits of integrating advanced AI-driven analyses into financial
strategies, offering a nuanced perspective on the interplay between digital
communication and market dynamics in an academic context. | q-fin.CP | Emoji Driven Crypto Assets Market Reactions |
2024-02-16T06:29:16Z | http://arxiv.org/pdf/2402.12170v2 | http://arxiv.org/abs/2402.12170v2 | Kuniaki Saito, Kihyuk Sohn, Chen-Yu Lee, Yoshitaka Ushiku | 2024-05-23T06:52:31Z | cs.CL, cs.AI | Large language models require updates to remain up-to-date or adapt to new
domains by fine-tuning them with new documents. One key is memorizing the
latest information in a way that the memorized information is extractable with
a query prompt. However, LLMs suffer from a phenomenon called perplexity curse;
despite minimizing document perplexity during fine-tuning, LLMs struggle to
extract information through a prompt sentence. In this new knowledge
acquisition and extraction, we find a very intriguing fact that LLMs can
accurately answer questions about the first sentence, but they struggle to
extract information described in the middle or end of the documents used for
fine-tuning. Our study suggests that the auto-regressive training causes this
issue; each token is prompted by reliance on all previous tokens, which hinders
the model from recalling information from training documents by question
prompts. To conduct the in-depth study, we publish both synthetic and real
datasets, enabling the evaluation of the QA performance w.r.t. the position of
the corresponding answer in a document. Our investigation shows that even a
large model suffers from the perplexity curse, but regularization such as
denoising auto-regressive loss can enhance the information extraction from
diverse positions. These findings will be (i) a key to improving knowledge
extraction from LLMs and (ii) new elements to discuss the trade-off between RAG
and fine-tuning in adapting LLMs to a new domain. | cs.CL | Where is the answer? Investigating Positional Bias in Language Model Knowledge Extraction |
2024-02-16T06:13:18Z | http://arxiv.org/pdf/2402.10466v4 | http://arxiv.org/abs/2402.10466v4 | Zekun Li, Zhiyu Zoey Chen, Mike Ross, Patrick Huber, Seungwhan Moon, Zhaojiang Lin, Xin Luna Dong, Adithya Sagar, Xifeng Yan, Paul A. Crook | 2024-05-30T04:19:54Z | cs.CL, cs.AI | Large language models (LLMs) are increasingly prevalent in conversational
systems due to their advanced understanding and generative capabilities in
general contexts. However, their effectiveness in task-oriented dialogues
(TOD), which requires not only response generation but also effective dialogue
state tracking (DST) within specific tasks and domains, remains less
satisfying. In this work, we propose a novel approach FnCTOD for solving DST
with LLMs through function calling. This method improves zero-shot DST,
allowing adaptation to diverse domains without extensive data collection or
model tuning. Our experimental results demonstrate that our approach achieves
exceptional performance with both modestly sized open-source and also
proprietary LLMs: with in-context prompting it enables various 7B or 13B
parameter models to surpass the previous state-of-the-art (SOTA) achieved by
ChatGPT, and improves ChatGPT's performance beating the SOTA by 5.6% average
joint goal accuracy (JGA). Individual model results for GPT-3.5 and GPT-4 are
boosted by 4.8% and 14%, respectively. We also show that by fine-tuning on a
small collection of diverse task-oriented dialogues, we can equip modestly
sized models, specifically a 13B parameter LLaMA2-Chat model, with
function-calling capabilities and DST performance comparable to ChatGPT while
maintaining their chat capabilities. We have made the code publicly available
at https://github.com/facebookresearch/FnCTOD | cs.CL | Large Language Models as Zero-shot Dialogue State Tracker through Function Calling |
2024-02-16T05:42:17Z | http://arxiv.org/pdf/2402.10462v1 | http://arxiv.org/abs/2402.10462v1 | Hossein Rajabzadeh, Mojtaba Valipour, Tianshu Zhu, Marzieh Tahaei, Hyock Ju Kwon, Ali Ghodsi, Boxing Chen, Mehdi Rezagholizadeh | 2024-02-16T05:42:17Z | cs.LG, cs.CL | Finetuning large language models requires huge GPU memory, restricting the
choice to acquire Larger models. While the quantized version of the Low-Rank
Adaptation technique, named QLoRA, significantly alleviates this issue, finding
the efficient LoRA rank is still challenging. Moreover, QLoRA is trained on a
pre-defined rank and, therefore, cannot be reconfigured for its lower ranks
without requiring further fine-tuning steps. This paper proposes QDyLoRA
-Quantized Dynamic Low-Rank Adaptation-, as an efficient quantization approach
for dynamic low-rank adaptation. Motivated by Dynamic LoRA, QDyLoRA is able to
efficiently finetune LLMs on a set of pre-defined LoRA ranks. QDyLoRA enables
fine-tuning Falcon-40b for ranks 1 to 64 on a single 32 GB V100-GPU through one
round of fine-tuning. Experimental results show that QDyLoRA is competitive to
QLoRA and outperforms when employing its optimal rank. | cs.LG | QDyLoRA: Quantized Dynamic Low-Rank Adaptation for Efficient Large Language Model Tuning |
2024-02-16T05:29:59Z | http://arxiv.org/pdf/2402.10987v2 | http://arxiv.org/abs/2402.10987v2 | Chenhui Hu, Pengfei Cao, Yubo Chen, Kang Liu, Jun Zhao | 2024-06-05T07:44:30Z | cs.CL, cs.AI | Knowledge editing aims to rectify inaccuracies in large language models
(LLMs) without costly retraining for outdated or erroneous knowledge. However,
current knowledge editing methods primarily focus on single editing, failing to
meet the requirements for lifelong editing. This study reveals a performance
degradation encountered by knowledge editing in lifelong editing, characterized
by toxicity buildup and toxicity flash, with the primary cause identified as
pattern unmatch. We introduce a knowledge editing approach named Wise-Layer
Knowledge Editor (WilKE), which selects editing layer based on the pattern
matching degree of editing knowledge across different layers in language
models. Experimental results demonstrate that, in lifelong editing, WilKE
exhibits an average improvement of 46.2% and 67.8% on editing GPT2-XL and GPT-J
relative to state-of-the-art knowledge editing methods. | cs.CL | WilKE: Wise-Layer Knowledge Editor for Lifelong Knowledge Editing |
2024-02-16T05:05:12Z | http://arxiv.org/pdf/2402.10986v3 | http://arxiv.org/abs/2402.10986v3 | Gagan Bhatia, El Moatez Billah Nagoudi, Hasan Cavusoglu, Muhammad Abdul-Mageed | 2024-06-14T13:26:47Z | cs.CL, cs.AI | We introduce FinTral, a suite of state-of-the-art multimodal large language
models (LLMs) built upon the Mistral-7b model and tailored for financial
analysis. FinTral integrates textual, numerical, tabular, and image data. We
enhance FinTral with domain-specific pretraining, instruction fine-tuning, and
RLAIF training by exploiting a large collection of textual and visual datasets
we curate for this work. We also introduce an extensive benchmark featuring
nine tasks and 25 datasets for evaluation, including hallucinations in the
financial domain. Our FinTral model trained with direct preference optimization
employing advanced Tools and Retrieval methods, dubbed FinTral-DPO-T&R,
demonstrates an exceptional zero-shot performance. It outperforms ChatGPT-3.5
in all tasks and surpasses GPT-4 in five out of nine tasks, marking a
significant advancement in AI-driven financial technology. We also demonstrate
that FinTral has the potential to excel in real-time analysis and
decision-making in diverse financial contexts. The GitHub repository for
FinTral is available at \url{https://github.com/UBC-NLP/fintral}. | cs.CL | FinTral: A Family of GPT-4 Level Multimodal Financial Large Language Models |
2024-02-16T05:03:01Z | http://arxiv.org/pdf/2402.10453v1 | http://arxiv.org/abs/2402.10453v1 | Navid Madani, Sougata Saha, Rohini Srihari | 2024-02-16T05:03:01Z | cs.CL | In this study, we address the challenge of consistently following emotional
support strategies in long conversations by large language models (LLMs). We
introduce the Strategy-Relevant Attention (SRA) metric, a model-agnostic
measure designed to evaluate the effectiveness of LLMs in adhering to strategic
prompts in emotional support contexts. By analyzing conversations within the
Emotional Support Conversations dataset (ESConv) using LLaMA models, we
demonstrate that SRA is significantly correlated with a model's ability to
sustain the outlined strategy throughout the interactions. Our findings reveal
that the application of SRA-informed prompts leads to enhanced strategic
adherence, resulting in conversations that more reliably exhibit the desired
emotional support strategies over longer conversations. Furthermore, we
contribute a comprehensive, multi-branch synthetic conversation dataset for
ESConv, featuring a variety of strategy continuations informed by our optimized
prompting method. The code and data are publicly available on our Github. | cs.CL | Steering Conversational Large Language Models for Long Emotional Support Conversations |
2024-02-16T04:41:33Z | http://arxiv.org/pdf/2402.10447v2 | http://arxiv.org/abs/2402.10447v2 | Shengjie Qiu, Junhao Zheng, Zhen Liu, Yicheng Luo, Qianli Ma | 2024-05-27T15:23:17Z | cs.CL, cs.LG | The incremental sequence labeling task involves continuously learning new
classes over time while retaining knowledge of the previous ones. Our
investigation identifies two significant semantic shifts: E2O (where the model
mislabels an old entity as a non-entity) and O2E (where the model labels a
non-entity or old entity as a new entity). Previous research has predominantly
focused on addressing the E2O problem, neglecting the O2E issue. This
negligence results in a model bias towards classifying new data samples as
belonging to the new class during the learning process. To address these
challenges, we propose a novel framework, Incremental Sequential Labeling
without Semantic Shifts (IS3). Motivated by the identified semantic shifts (E2O
and O2E), IS3 aims to mitigate catastrophic forgetting in models. As for the
E2O problem, we use knowledge distillation to maintain the model's
discriminative ability for old entities. Simultaneously, to tackle the O2E
problem, we alleviate the model's bias towards new entities through debiased
loss and optimization levels. Our experimental evaluation, conducted on three
datasets with various incremental settings, demonstrates the superior
performance of IS3 compared to the previous state-of-the-art method by a
significant margin.The data, code, and scripts are publicly available at
https://github.com/zzz47zzz/codebase-for-incremental-learning-with-llm. | cs.CL | Incremental Sequence Labeling: A Tale of Two Shifts |
2024-02-16T03:54:48Z | http://arxiv.org/pdf/2402.10436v1 | http://arxiv.org/abs/2402.10436v1 | Wenchao Dong, Assem Zhunis, Hyojin Chin, Jiyoung Han, Meeyoung Cha | 2024-02-16T03:54:48Z | cs.CL | We explored cultural biases-individualism vs. collectivism-in ChatGPT across
three Western languages (i.e., English, German, and French) and three Eastern
languages (i.e., Chinese, Japanese, and Korean). When ChatGPT adopted an
individualistic persona in Western languages, its collectivism scores (i.e.,
out-group values) exhibited a more negative trend, surpassing their positive
orientation towards individualism (i.e., in-group values). Conversely, when a
collectivistic persona was assigned to ChatGPT in Eastern languages, a similar
pattern emerged with more negative responses toward individualism (i.e.,
out-group values) as compared to collectivism (i.e., in-group values). The
results indicate that when imbued with a particular social identity, ChatGPT
discerns in-group and out-group, embracing in-group values while eschewing
out-group values. Notably, the negativity towards the out-group, from which
prejudices and discrimination arise, exceeded the positivity towards the
in-group. The experiment was replicated in the political domain, and the
results remained consistent. Furthermore, this replication unveiled an
intrinsic Democratic bias in Large Language Models (LLMs), aligning with
earlier findings and providing integral insights into mitigating such bias
through prompt engineering. Extensive robustness checks were performed using
varying hyperparameter and persona setup methods, with or without social
identity labels, across other popular language models. | cs.CL | I Am Not Them: Fluid Identities and Persistent Out-group Bias in Large Language Models |
2024-02-16T03:39:37Z | http://arxiv.org/pdf/2402.10430v1 | http://arxiv.org/abs/2402.10430v1 | Dheeraj Mekala, Alex Nguyen, Jingbo Shang | 2024-02-16T03:39:37Z | cs.CL | Instruction-tuning language models has become a crucial step in aligning them
for general use. Typically, this process involves extensive training on large
datasets, incurring high training costs. In this paper, we introduce a novel
training data selection based on the learning percentage of the samples. We
assert that current language models possess the capability to autonomously
select high-quality training data, leading to comparable or improved
performance compared to training on the entire dataset. Our experiments span
different-sized models, revealing that this characteristic holds for models
ranging from 1B (small) to 13B (large) in size. Moreover, we demonstrate an
interesting finding that the data hardness transfers across model sizes, and a
smaller 350M model can effectively curate high-quality training data with hard
samples for a larger 13B model, resulting in an equally or superior
instruction-tuned model compared to training on the complete dataset. Utilizing
open-sourced OPT and Llama-2 models up to 13B in size, two publicly available
instruction-tuning training datasets and evaluated by both automatic metrics &
humans, our paper introduces a novel approach to training data selection,
showcasing a more efficient alternative. | cs.CL | Smaller Language Models are capable of selecting Instruction-Tuning Training Data for Larger Language Models |
2024-02-16T03:30:27Z | http://arxiv.org/pdf/2402.10427v1 | http://arxiv.org/abs/2402.10427v1 | Muqiao Yang, Xiang Li, Umberto Cappellazzo, Shinji Watanabe, Bhiksha Raj | 2024-02-16T03:30:27Z | cs.CL, cs.AI, cs.SD, eess.AS | Continual learning has emerged as an increasingly important challenge across
various tasks, including Spoken Language Understanding (SLU). In SLU, its
objective is to effectively handle the emergence of new concepts and evolving
environments. The evaluation of continual learning algorithms typically
involves assessing the model's stability, plasticity, and generalizability as
fundamental aspects of standards. However, existing continual learning metrics
primarily focus on only one or two of the properties. They neglect the overall
performance across all tasks, and do not adequately disentangle the plasticity
versus stability/generalizability trade-offs within the model. In this work, we
propose an evaluation methodology that provides a unified evaluation on
stability, plasticity, and generalizability in continual learning. By employing
the proposed metric, we demonstrate how introducing various knowledge
distillations can improve different aspects of these three properties of the
SLU model. We further show that our proposed metric is more sensitive in
capturing the impact of task ordering in continual learning, making it better
suited for practical use-case scenarios. | cs.CL | Evaluating and Improving Continual Learning in Spoken Language Understanding |
2024-02-16T03:24:56Z | http://arxiv.org/pdf/2402.10426v1 | http://arxiv.org/abs/2402.10426v1 | Herun Wan, Shangbin Feng, Zhaoxuan Tan, Heng Wang, Yulia Tsvetkov, Minnan Luo | 2024-02-16T03:24:56Z | cs.CL | Large language models are limited by challenges in factuality and
hallucinations to be directly employed off-the-shelf for judging the veracity
of news articles, where factual accuracy is paramount. In this work, we propose
DELL that identifies three key stages in misinformation detection where LLMs
could be incorporated as part of the pipeline: 1) LLMs could \emph{generate
news reactions} to represent diverse perspectives and simulate user-news
interaction networks; 2) LLMs could \emph{generate explanations} for proxy
tasks (e.g., sentiment, stance) to enrich the contexts of news articles and
produce experts specializing in various aspects of news understanding; 3) LLMs
could \emph{merge task-specific experts} and provide an overall prediction by
incorporating the predictions and confidence scores of varying experts.
Extensive experiments on seven datasets with three LLMs demonstrate that DELL
outperforms state-of-the-art baselines by up to 16.8\% in macro f1-score.
Further analysis reveals that the generated reactions and explanations are
greatly helpful in misinformation detection, while our proposed LLM-guided
expert merging helps produce better-calibrated predictions. | cs.CL | DELL: Generating Reactions and Explanations for LLM-Based Misinformation Detection |
2024-02-16T03:20:14Z | http://arxiv.org/pdf/2402.10424v1 | http://arxiv.org/abs/2402.10424v1 | Ting-Rui Chiang, Dani Yogatama | 2024-02-16T03:20:14Z | cs.CL, cs.AI | Many existing theoretical analyses of in-context learning for natural
language processing are based on latent variable models that leaves gaps
between theory and practice. We aim to close these gaps by proposing a
theoretical framework, the Pelican Soup Framework. In this framework, we
introduce (1) the notion of a common sense knowledge base, (2) a general
formalism for natural language classification tasks, and the notion of (3)
meaning association. Under this framework, we can establish a
$\mathcal{O}(1/T)$ loss bound for in-context learning, where $T$ is the number
of example-label pairs in the demonstration. Compared with previous works, our
bound reflects the effect of the choice of verbalizers and the effect of
instruction tuning. An additional notion of \textit{atom concepts} makes our
framework possible to explain the generalization to tasks unseen in the
language model training data. Finally, we propose a toy setup, Calcutec, and a
digit addition task that mimics types of distribution shifts a model needs to
overcome to perform in-context learning. We also experiment with GPT2-Large on
real-world NLP tasks. Our empirical results demonstrate the efficacy of our
framework to explain in-context learning. | cs.CL | Understanding In-Context Learning with a Pelican Soup Framework |
2024-02-16T03:06:37Z | http://arxiv.org/pdf/2402.10422v2 | http://arxiv.org/abs/2402.10422v2 | Ioannis Tsiamas, Gerard I. Gállego, José A. R. Fonollosa, Marta R. Costa-jussà | 2024-06-05T20:56:20Z | cs.CL | Data scarcity and the modality gap between the speech and text modalities are
two major obstacles of end-to-end Speech Translation (ST) systems, thus
hindering their performance. Prior work has attempted to mitigate these
challenges by leveraging external MT data and optimizing distance metrics that
bring closer the speech-text representations. However, achieving competitive
results typically requires some ST data. For this reason, we introduce
ZeroSwot, a method for zero-shot ST that bridges the modality gap without any
paired ST data. Leveraging a novel CTC compression and Optimal Transport, we
train a speech encoder using only ASR data, to align with the representation
space of a massively multilingual MT model. The speech encoder seamlessly
integrates with the MT model at inference, enabling direct translation from
speech to text, across all languages supported by the MT model. Our experiments
show that we can effectively close the modality gap without ST data, while our
results on MuST-C and CoVoST demonstrate our method's superiority over not only
previous zero-shot models, but also supervised ones, achieving state-of-the-art
results. | cs.CL | Pushing the Limits of Zero-shot End-to-End Speech Translation |
2024-02-16T02:47:09Z | http://arxiv.org/pdf/2402.10416v1 | http://arxiv.org/abs/2402.10416v1 | Lance Ying, Tan Zhi-Xuan, Lionel Wong, Vikash Mansinghka, Joshua Tenenbaum | 2024-02-16T02:47:09Z | cs.AI, cs.CL | Despite the fact that beliefs are mental states that cannot be directly
observed, humans talk about each others' beliefs on a regular basis, often
using rich compositional language to describe what others think and know. What
explains this capacity to interpret the hidden epistemic content of other
minds? In this paper, we take a step towards an answer by grounding the
semantics of belief statements in a Bayesian theory-of-mind: By modeling how
humans jointly infer coherent sets of goals, beliefs, and plans that explain an
agent's actions, then evaluating statements about the agent's beliefs against
these inferences via epistemic logic, our framework provides a conceptual role
semantics for belief, explaining the gradedness and compositionality of human
belief attributions, as well as their intimate connection with goals and plans.
We evaluate this framework by studying how humans attribute goals and beliefs
while watching an agent solve a doors-and-keys gridworld puzzle that requires
instrumental reasoning about hidden objects. In contrast to pure logical
deduction, non-mentalizing baselines, and mentalizing that ignores the role of
instrumental plans, our model provides a much better fit to human goal and
belief attributions, demonstrating the importance of theory-of-mind for a
semantics of belief. | cs.AI | Grounding Language about Belief in a Bayesian Theory-of-Mind |
2024-02-16T02:32:06Z | http://arxiv.org/pdf/2402.10412v2 | http://arxiv.org/abs/2402.10412v2 | Jiaheng Wei, Yuanshun Yao, Jean-Francois Ton, Hongyi Guo, Andrew Estornell, Yang Liu | 2024-06-06T18:33:02Z | cs.CL, cs.AI, cs.LG | LLM hallucination, i.e. generating factually incorrect yet seemingly
convincing answers, is currently a major threat to the trustworthiness and
reliability of LLMs. The first step towards solving this complicated problem is
to measure it. However, existing hallucination metrics require having a
benchmark dataset with gold-standard answers, i.e. "best" or "correct" answers
written by humans. Such requirements make hallucination measurement costly and
prone to human errors. In this work, we propose Factualness Evaluations via
Weighting LLMs (FEWL), an innovative hallucination metric that is specifically
designed for the scenario when gold-standard answers are absent. FEWL leverages
the answers from off-the-shelf LLMs that serve as a proxy of gold-standard
answers. The key challenge is how to quantify the expertise of reference LLMs
resourcefully. We show FEWL has certain theoretical guarantees and demonstrate
empirically it gives more accurate hallucination measures than naively using
reference LLMs. We also show how to leverage FEWL to reduce hallucination
through both in-context learning and supervised fine-tuning. Extensive
experiment results on Truthful-QA, CHALE, and HaluEval datasets demonstrate the
effectiveness of FEWL. | cs.CL | Measuring and Reducing LLM Hallucination without Gold-Standard Answers |
2024-02-16T02:21:59Z | http://arxiv.org/pdf/2402.10409v1 | http://arxiv.org/abs/2402.10409v1 | Jun Zhuang, Casey Kennington | 2024-02-16T02:21:59Z | cs.CL, cs.AI, cs.IR, cs.LG | As new research on Large Language Models (LLMs) continues, it is difficult to
keep up with new research and models. To help researchers synthesize the new
research many have written survey papers, but even those have become numerous.
In this paper, we develop a method to automatically assign survey papers to a
taxonomy. We collect the metadata of 144 LLM survey papers and explore three
paradigms to classify papers within the taxonomy. Our work indicates that
leveraging graph structure information on co-category graphs can significantly
outperform the language models in two paradigms; pre-trained language models'
fine-tuning and zero-shot/few-shot classifications using LLMs. We find that our
model surpasses an average human recognition level and that fine-tuning LLMs
using weak labels generated by a smaller model, such as the GCN in this study,
can be more effective than using ground-truth labels, revealing the potential
of weak-to-strong generalization in the taxonomy classification task. | cs.CL | Understanding Survey Paper Taxonomy about Large Language Models via Graph Representation Learning |
2024-02-16T01:54:43Z | http://arxiv.org/pdf/2402.10400v2 | http://arxiv.org/abs/2402.10400v2 | Sergio Servantez, Joe Barrow, Kristian Hammond, Rajiv Jain | 2024-02-23T18:55:23Z | cs.CL | Rule-based reasoning, a fundamental type of legal reasoning, enables us to
draw conclusions by accurately applying a rule to a set of facts. We explore
causal language models as rule-based reasoners, specifically with respect to
compositional rules - rules consisting of multiple elements which form a
complex logical expression. Reasoning about compositional rules is challenging
because it requires multiple reasoning steps, and attending to the logical
relationships between elements. We introduce a new prompting method, Chain of
Logic, which elicits rule-based reasoning through decomposition (solving
elements as independent threads of logic), and recomposition (recombining these
sub-answers to resolve the underlying logical expression). This method was
inspired by the IRAC (Issue, Rule, Application, Conclusion) framework, a
sequential reasoning approach used by lawyers. We evaluate chain of logic
across eight rule-based reasoning tasks involving three distinct compositional
rules from the LegalBench benchmark and demonstrate it consistently outperforms
other prompting methods, including chain of thought and self-ask, using
open-source and commercial language models. | cs.CL | Chain of Logic: Rule-Based Reasoning with Large Language Models |
2024-02-16T00:25:24Z | http://arxiv.org/pdf/2402.10380v1 | http://arxiv.org/abs/2402.10380v1 | Junhyun Lee, Wooseong Yang, Jaewoo Kang | 2024-02-16T00:25:24Z | cs.LG, cs.AI, cs.CL | In the evolving landscape of machine learning, the adaptation of pre-trained
models through prompt tuning has become increasingly prominent. This trend is
particularly observable in the graph domain, where diverse pre-training
strategies present unique challenges in developing effective prompt-based
tuning methods for graph neural networks. Previous approaches have been
limited, focusing on specialized prompting functions tailored to models with
edge prediction pre-training tasks. These methods, however, suffer from a lack
of generalizability across different pre-training strategies. Recently, a
simple prompt tuning method has been designed for any pre-training strategy,
functioning within the input graph's feature space. This allows it to
theoretically emulate any type of prompting function, thereby significantly
increasing its versatility for a range of downstream applications.
Nevertheless, the capacity of such simple prompts to fully grasp the complex
contexts found in graphs remains an open question, necessitating further
investigation. Addressing this challenge, our work introduces the
Subgraph-level Universal Prompt Tuning (SUPT) approach, focusing on the
detailed context within subgraphs. In SUPT, prompt features are assigned at the
subgraph-level, preserving the method's universal capability. This requires
extremely fewer tuning parameters than fine-tuning-based methods, outperforming
them in 42 out of 45 full-shot scenario experiments with an average improvement
of over 2.5%. In few-shot scenarios, it excels in 41 out of 45 experiments,
achieving an average performance increase of more than 6.6%. | cs.LG | Subgraph-level Universal Prompt Tuning |
2024-02-16T00:10:26Z | http://arxiv.org/pdf/2402.10379v2 | http://arxiv.org/abs/2402.10379v2 | Ajay Patel, Colin Raffel, Chris Callison-Burch | 2024-05-27T19:54:44Z | cs.CL, cs.LG | Large language models (LLMs) have become a dominant and important tool for
NLP researchers in a wide range of tasks. Today, many researchers use LLMs in
synthetic data generation, task evaluation, fine-tuning, distillation, and
other model-in-the-loop research workflows. However, challenges arise when
using these models that stem from their scale, their closed source nature, and
the lack of standardized tooling for these new and emerging workflows. The
rapid rise to prominence of these models and these unique challenges has had
immediate adverse impacts on open science and on the reproducibility of work
that uses them. In this paper, we introduce DataDreamer, an open source Python
library that allows researchers to write simple code to implement powerful LLM
workflows. DataDreamer also helps researchers adhere to best practices that we
propose to encourage open science and reproducibility. The library and
documentation are available at https://github.com/datadreamer-dev/DataDreamer . | cs.CL | DataDreamer: A Tool for Synthetic Data Generation and Reproducible LLM Workflows |
2024-02-15T23:39:04Z | http://arxiv.org/pdf/2402.10373v2 | http://arxiv.org/abs/2402.10373v2 | Yanis Labrak, Adrien Bazoge, Emmanuel Morin, Pierre-Antoine Gourraud, Mickael Rouvier, Richard Dufour | 2024-06-09T15:19:09Z | cs.CL, cs.AI, cs.LG | Large Language Models (LLMs) have demonstrated remarkable versatility in
recent years, offering potential applications across specialized domains such
as healthcare and medicine. Despite the availability of various open-source
LLMs tailored for health contexts, adapting general-purpose LLMs to the medical
domain presents significant challenges. In this paper, we introduce BioMistral,
an open-source LLM tailored for the biomedical domain, utilizing Mistral as its
foundation model and further pre-trained on PubMed Central. We conduct a
comprehensive evaluation of BioMistral on a benchmark comprising 10 established
medical question-answering (QA) tasks in English. We also explore lightweight
models obtained through quantization and model merging approaches. Our results
demonstrate BioMistral's superior performance compared to existing open-source
medical models and its competitive edge against proprietary counterparts.
Finally, to address the limited availability of data beyond English and to
assess the multilingual generalization of medical LLMs, we automatically
translated and evaluated this benchmark into 7 other languages. This marks the
first large-scale multilingual evaluation of LLMs in the medical domain.
Datasets, multilingual evaluation benchmarks, scripts, and all the models
obtained during our experiments are freely released. | cs.CL | BioMistral: A Collection of Open-Source Pretrained Large Language Models for Medical Domains |
2024-02-15T23:09:42Z | http://arxiv.org/pdf/2402.10359v2 | http://arxiv.org/abs/2402.10359v2 | Zheyuan Liu, Xiaoxin He, Yijun Tian, Nitesh V. Chawla | 2024-03-16T19:56:04Z | cs.LG, cs.CL | Graph plays an important role in representing complex relationships in
real-world applications such as social networks, biological data and citation
networks. In recent years, Large Language Models (LLMs) have achieved
tremendous success in various domains, which makes applying LLMs to graphs
particularly appealing. However, directly applying LLMs to graph modalities
presents unique challenges due to the discrepancy and mismatch between the
graph and text modalities. Hence, to further investigate LLMs' potential for
comprehending graph information, we introduce GraphPrompter, a novel framework
designed to align graph information with LLMs via soft prompts. Specifically,
GraphPrompter consists of two main components: a graph neural network to encode
complex graph information and an LLM that effectively processes textual
information. Comprehensive experiments on various benchmark datasets under node
classification and link prediction tasks demonstrate the effectiveness of our
proposed method. The GraphPrompter framework unveils the substantial
capabilities of LLMs as predictors in graph-related tasks, enabling researchers
to utilize LLMs across a spectrum of real-world graph scenarios more
effectively. | cs.LG | Can we Soft Prompt LLMs for Graph Learning Tasks? |
2024-02-15T22:54:24Z | http://arxiv.org/pdf/2402.10353v1 | http://arxiv.org/abs/2402.10353v1 | Kang He, Yinghan Long, Kaushik Roy | 2024-02-15T22:54:24Z | cs.CL, cs.LG | Prompt learning is susceptible to intrinsic bias present in pre-trained
language models (LMs), resulting in sub-optimal performance of prompt-based
zero/few-shot learning. In this work, we propose a null-input prompting method
to calibrate intrinsic bias encoded in pre-trained LMs. Different from prior
efforts that address intrinsic bias primarily for social fairness and often
involve excessive computational cost, our objective is to explore enhancing
LMs' performance in downstream zero/few-shot learning while emphasizing the
efficiency of intrinsic bias calibration. Specifically, we leverage a diverse
set of auto-selected null-meaning inputs generated from GPT-4 to prompt
pre-trained LMs for intrinsic bias probing. Utilizing the bias-reflected
probability distribution, we formulate a distribution disparity loss for bias
calibration, where we exclusively update bias parameters ($0.1\%$ of total
parameters) of LMs towards equal probability distribution. Experimental results
show that the calibration promotes an equitable starting point for LMs while
preserving language modeling abilities. Across a wide range of datasets,
including sentiment analysis and topic classification, our method significantly
improves zero/few-shot learning performance of LMs for both in-context learning
and prompt-based fine-tuning (on average $9\%$ and $2\%$, respectively). | cs.CL | Prompt-Based Bias Calibration for Better Zero/Few-Shot Learning of Language Models |
2024-02-15T22:19:41Z | http://arxiv.org/pdf/2403.19669v2 | http://arxiv.org/abs/2403.19669v2 | Allison Chen, Ilia Sucholutsky, Olga Russakovsky, Thomas L. Griffiths | 2024-05-10T17:33:24Z | cs.LG, cs.AI, cs.CL, cs.CV | Does language help make sense of the visual world? How important is it to
actually see the world rather than having it described with words? These basic
questions about the nature of intelligence have been difficult to answer
because we only had one example of an intelligent system -- humans -- and
limited access to cases that isolated language or vision. However, the
development of sophisticated Vision-Language Models (VLMs) by artificial
intelligence researchers offers us new opportunities to explore the
contributions that language and vision make to learning about the world. We
ablate components from the cognitive architecture of these models to identify
their contributions to learning new tasks from limited data. We find that a
language model leveraging all components recovers a majority of a VLM's
performance, despite its lack of visual input, and that language seems to allow
this by providing access to prior knowledge and reasoning. | cs.LG | Analyzing the Roles of Language and Vision in Learning from Limited Data |
2024-02-15T22:11:18Z | http://arxiv.org/pdf/2402.10342v1 | http://arxiv.org/abs/2402.10342v1 | Yihan Du, Anna Winnicki, Gal Dalal, Shie Mannor, R. Srikant | 2024-02-15T22:11:18Z | cs.LG, cs.AI, cs.CL | Reinforcement Learning from Human Feedback (RLHF) has achieved impressive
empirical successes while relying on a small amount of human feedback. However,
there is limited theoretical justification for this phenomenon. Additionally,
most recent studies focus on value-based algorithms despite the recent
empirical successes of policy-based algorithms. In this work, we consider an
RLHF algorithm based on policy optimization (PO-RLHF). The algorithm is based
on the popular Policy Cover-Policy Gradient (PC-PG) algorithm, which assumes
knowledge of the reward function. In PO-RLHF, knowledge of the reward function
is not assumed and the algorithm relies on trajectory-based comparison feedback
to infer the reward function. We provide performance bounds for PO-RLHF with
low query complexity, which provides insight into why a small amount of human
feedback may be sufficient to get good performance with RLHF. A key novelty is
our trajectory-level elliptical potential analysis technique used to infer
reward function parameters when comparison queries rather than reward
observations are used. We provide and analyze algorithms in two settings:
linear and neural function approximation, PG-RLHF and NN-PG-RLHF, respectively. | cs.LG | Exploration-Driven Policy Optimization in RLHF: Theoretical Insights on Efficient Data Utilization |
2024-02-15T20:26:07Z | http://arxiv.org/pdf/2402.10979v2 | http://arxiv.org/abs/2402.10979v2 | Yebowen Hu, Kaiqiang Song, Sangwoo Cho, Xiaoyang Wang, Hassan Foroosh, Dong Yu, Fei Liu | 2024-06-16T06:43:50Z | cs.CL, cs.AI | Large language models hold significant potential for integrating various data
types, such as text documents and database records, for advanced analytics.
However, blending text and numerical data presents substantial challenges. LLMs
need to process and cross-reference entities and numbers, handle data
inconsistencies and redundancies, and develop planning capabilities such as
building a working memory for managing complex data queries. In this paper, we
introduce four novel tasks centered around sports data analytics to evaluate
the numerical reasoning and information fusion capabilities of LLMs. These
tasks involve providing LLMs with detailed, play-by-play sports game
descriptions, then challenging them with adversarial scenarios such as new game
rules, longer durations, scrambled narratives, and analyzing key statistics in
game summaries. We conduct extensive experiments on NBA and NFL games to assess
the performance of LLMs on these tasks. Our benchmark, SportsMetrics,
introduces a new mechanism for assessing LLMs' numerical reasoning and fusion
skills. | cs.CL | SportsMetrics: Blending Text and Numerical Data to Understand Information Fusion in LLMs |
2024-02-15T20:24:39Z | http://arxiv.org/pdf/2402.10311v5 | http://arxiv.org/abs/2402.10311v5 | Ramon Ferrer-i-Cancho | 2024-03-22T16:10:14Z | cs.CL, physics.soc-ph | The word order of a sentence is shaped by multiple principles. The principle
of syntactic dependency distance minimization is in conflict with the principle
of surprisal minimization (or predictability maximization) in single head
syntactic dependency structures: while the former predicts that the head should
be placed at the center of the linear arrangement, the latter predicts that the
head should be placed at one of the ends (either first or last). A critical
question is when surprisal minimization (or predictability maximization) should
surpass syntactic dependency distance minimization. In the context of single
head structures, it has been predicted that this is more likely to happen when
two conditions are met, i.e. (a) fewer words are involved and (b) words are
shorter. Here we test the prediction on the noun phrase when it is composed of
a demonstrative, a numeral, an adjective and a noun. We find that, across
preferred orders in languages, the noun tends to be placed at one of the ends,
confirming the theoretical prediction. We also show evidence of anti locality
effects: syntactic dependency distances in preferred orders are longer than
expected by chance. | cs.CL | The optimal placement of the head in the noun phrase. The case of demonstrative, numeral, adjective and noun |
2024-02-15T20:08:07Z | http://arxiv.org/pdf/2402.10302v1 | http://arxiv.org/abs/2402.10302v1 | Oleg Vasilyev, John Bohannon | 2024-02-15T20:08:07Z | cs.CL | We found that a simple property of clusters in a clustered dataset of news
correlate strongly with importance and urgency of news (IUN) as assessed by
LLM. We verified our finding across different news datasets, dataset sizes,
clustering algorithms and embeddings. The found correlation should allow using
clustering (as an alternative to LLM) for identifying the most important urgent
news, or for filtering out unimportant articles. | cs.CL | How to Discern Important Urgent News? |
2024-02-15T19:53:11Z | http://arxiv.org/pdf/2402.10294v1 | http://arxiv.org/abs/2402.10294v1 | Bryan Wang, Yuliang Li, Zhaoyang Lv, Haijun Xia, Yan Xu, Raj Sodhi | 2024-02-15T19:53:11Z | cs.HC, cs.AI, cs.CL, cs.MM | Video creation has become increasingly popular, yet the expertise and effort
required for editing often pose barriers to beginners. In this paper, we
explore the integration of large language models (LLMs) into the video editing
workflow to reduce these barriers. Our design vision is embodied in LAVE, a
novel system that provides LLM-powered agent assistance and language-augmented
editing features. LAVE automatically generates language descriptions for the
user's footage, serving as the foundation for enabling the LLM to process
videos and assist in editing tasks. When the user provides editing objectives,
the agent plans and executes relevant actions to fulfill them. Moreover, LAVE
allows users to edit videos through either the agent or direct UI manipulation,
providing flexibility and enabling manual refinement of agent actions. Our user
study, which included eight participants ranging from novices to proficient
editors, demonstrated LAVE's effectiveness. The results also shed light on user
perceptions of the proposed LLM-assisted editing paradigm and its impact on
users' creativity and sense of co-creation. Based on these findings, we propose
design implications to inform the future development of agent-assisted content
editing. | cs.HC | LAVE: LLM-Powered Agent Assistance and Language Augmentation for Video Editing |
2024-02-15T18:59:18Z | http://arxiv.org/pdf/2402.10210v1 | http://arxiv.org/abs/2402.10210v1 | Huizhuo Yuan, Zixiang Chen, Kaixuan Ji, Quanquan Gu | 2024-02-15T18:59:18Z | cs.LG, cs.AI, cs.CL, cs.CV, stat.ML | Fine-tuning Diffusion Models remains an underexplored frontier in generative
artificial intelligence (GenAI), especially when compared with the remarkable
progress made in fine-tuning Large Language Models (LLMs). While cutting-edge
diffusion models such as Stable Diffusion (SD) and SDXL rely on supervised
fine-tuning, their performance inevitably plateaus after seeing a certain
volume of data. Recently, reinforcement learning (RL) has been employed to
fine-tune diffusion models with human preference data, but it requires at least
two images ("winner" and "loser" images) for each text prompt. In this paper,
we introduce an innovative technique called self-play fine-tuning for diffusion
models (SPIN-Diffusion), where the diffusion model engages in competition with
its earlier versions, facilitating an iterative self-improvement process. Our
approach offers an alternative to conventional supervised fine-tuning and RL
strategies, significantly improving both model performance and alignment. Our
experiments on the Pick-a-Pic dataset reveal that SPIN-Diffusion outperforms
the existing supervised fine-tuning method in aspects of human preference
alignment and visual appeal right from its first iteration. By the second
iteration, it exceeds the performance of RLHF-based methods across all metrics,
achieving these results with less data. | cs.LG | Self-Play Fine-Tuning of Diffusion Models for Text-to-Image Generation |
2024-02-15T18:59:02Z | http://arxiv.org/pdf/2402.10208v2 | http://arxiv.org/abs/2402.10208v2 | Eliahu Horwitz, Jonathan Kahana, Yedid Hoshen | 2024-07-01T12:48:51Z | cs.LG, cs.CL, cs.CR, cs.CV | The dominant paradigm in generative modeling consists of two steps: i)
pre-training on a large-scale but unsafe dataset, ii) aligning the pre-trained
model with human values via fine-tuning. This practice is considered safe, as
no current method can recover the unsafe, pre-fine-tuning model weights. In
this paper, we demonstrate that this assumption is often false. Concretely, we
present Spectral DeTuning, a method that can recover the weights of the
pre-fine-tuning model using a few low-rank (LoRA) fine-tuned models. In
contrast to previous attacks that attempt to recover pre-fine-tuning
capabilities, our method aims to recover the exact pre-fine-tuning weights. Our
approach exploits this new vulnerability against large-scale models such as a
personalized Stable Diffusion and an aligned Mistral. | cs.LG | Recovering the Pre-Fine-Tuning Weights of Generative Models |
2024-02-15T18:58:31Z | http://arxiv.org/pdf/2402.10207v5 | http://arxiv.org/abs/2402.10207v5 | Rui Yang, Xiaoman Pan, Feng Luo, Shuang Qiu, Han Zhong, Dong Yu, Jianshu Chen | 2024-06-05T09:25:41Z | cs.LG, cs.AI, cs.CL | We consider the problem of multi-objective alignment of foundation models
with human preferences, which is a critical step towards helpful and harmless
AI systems. However, it is generally costly and unstable to fine-tune large
foundation models using reinforcement learning (RL), and the
multi-dimensionality, heterogeneity, and conflicting nature of human
preferences further complicate the alignment process. In this paper, we
introduce Rewards-in-Context (RiC), which conditions the response of a
foundation model on multiple rewards in its prompt context and applies
supervised fine-tuning for alignment. The salient features of RiC are
simplicity and adaptivity, as it only requires supervised fine-tuning of a
single foundation model and supports dynamic adjustment for user preferences
during inference time. Inspired by the analytical solution of an abstracted
convex optimization problem, our dynamic inference-time adjustment method
approaches the Pareto-optimal solution for multiple objectives. Empirical
evidence demonstrates the efficacy of our method in aligning both Large
Language Models (LLMs) and diffusion models to accommodate diverse rewards with
only around 10% GPU hours compared with multi-objective RL baseline. | cs.LG | Rewards-in-Context: Multi-objective Alignment of Foundation Models with Dynamic Preference Adjustment |
2024-02-15T18:58:09Z | http://arxiv.org/pdf/2402.10260v1 | http://arxiv.org/abs/2402.10260v1 | Alexandra Souly, Qingyuan Lu, Dillon Bowen, Tu Trinh, Elvis Hsieh, Sana Pandey, Pieter Abbeel, Justin Svegliato, Scott Emmons, Olivia Watkins, Sam Toyer | 2024-02-15T18:58:09Z | cs.LG, cs.CL, cs.CR | The rise of large language models (LLMs) has drawn attention to the existence
of "jailbreaks" that allow the models to be used maliciously. However, there is
no standard benchmark for measuring the severity of a jailbreak, leaving
authors of jailbreak papers to create their own. We show that these benchmarks
often include vague or unanswerable questions and use grading criteria that are
biased towards overestimating the misuse potential of low-quality model
responses. Some jailbreak techniques make the problem worse by decreasing the
quality of model responses even on benign questions: we show that several
jailbreaking techniques substantially reduce the zero-shot performance of GPT-4
on MMLU. Jailbreaks can also make it harder to elicit harmful responses from an
"uncensored" open-source model. We present a new benchmark, StrongREJECT, which
better discriminates between effective and ineffective jailbreaks by using a
higher-quality question set and a more accurate response grading algorithm. We
show that our new grading scheme better accords with human judgment of response
quality and overall jailbreak effectiveness, especially on the sort of
low-quality responses that contribute the most to over-estimation of jailbreak
performance on existing benchmarks. We release our code and data at
https://github.com/alexandrasouly/strongreject. | cs.LG | A StrongREJECT for Empty Jailbreaks |
2024-02-15T18:55:41Z | http://arxiv.org/pdf/2402.10200v2 | http://arxiv.org/abs/2402.10200v2 | Xuezhi Wang, Denny Zhou | 2024-05-23T20:53:59Z | cs.CL | In enhancing the reasoning capabilities of large language models (LLMs),
prior research primarily focuses on specific prompting techniques such as
few-shot or zero-shot chain-of-thought (CoT) prompting. These methods, while
effective, often involve manually intensive prompt engineering. Our study takes
a novel approach by asking: Can LLMs reason effectively without prompting? Our
findings reveal that, intriguingly, CoT reasoning paths can be elicited from
pre-trained LLMs by simply altering the \textit{decoding} process. Rather than
conventional greedy decoding, we investigate the top-$k$ alternative tokens,
uncovering that CoT paths are frequently inherent in these sequences. This
approach not only bypasses the confounders of prompting but also allows us to
assess the LLMs' \textit{intrinsic} reasoning abilities. Moreover, we observe
that the presence of a CoT in the decoding path correlates with a higher
confidence in the model's decoded answer. This confidence metric effectively
differentiates between CoT and non-CoT paths. Extensive empirical studies on
various reasoning benchmarks show that the proposed CoT-decoding effectively
elicits reasoning capabilities from language models, which were previously
obscured by standard greedy decoding. | cs.CL | Chain-of-Thought Reasoning Without Prompting |
2024-02-15T18:51:32Z | http://arxiv.org/pdf/2402.10196v1 | http://arxiv.org/abs/2402.10196v1 | Lingbo Mo, Zeyi Liao, Boyuan Zheng, Yu Su, Chaowei Xiao, Huan Sun | 2024-02-15T18:51:32Z | cs.CL, cs.AI | Language agents powered by large language models (LLMs) have seen exploding
development. Their capability of using language as a vehicle for thought and
communication lends an incredible level of flexibility and versatility. People
have quickly capitalized on this capability to connect LLMs to a wide range of
external components and environments: databases, tools, the Internet, robotic
embodiment, etc. Many believe an unprecedentedly powerful automation technology
is emerging. However, new automation technologies come with new safety risks,
especially for intricate systems like language agents. There is a surprisingly
large gap between the speed and scale of their development and deployment and
our understanding of their safety risks. Are we building a house of cards? In
this position paper, we present the first systematic effort in mapping
adversarial attacks against language agents. We first present a unified
conceptual framework for agents with three major components: Perception, Brain,
and Action. Under this framework, we present a comprehensive discussion and
propose 12 potential attack scenarios against different components of an agent,
covering different attack strategies (e.g., input manipulation, adversarial
demonstrations, jailbreaking, backdoors). We also draw connections to
successful attack strategies previously applied to LLMs. We emphasize the
urgency to gain a thorough understanding of language agent risks before their
widespread deployment. | cs.CL | A Trembling House of Cards? Mapping Adversarial Attacks against Language Agents |
2024-02-15T18:50:06Z | http://arxiv.org/pdf/2402.10193v2 | http://arxiv.org/abs/2402.10193v2 | James Liu, Guangxuan Xiao, Kai Li, Jason D. Lee, Song Han, Tri Dao, Tianle Cai | 2024-02-28T03:42:10Z | cs.LG, cs.CL | Large Language Models (LLMs) are typically trained in two phases:
pre-training on large internet-scale datasets, and fine-tuning for downstream
tasks. Given the higher computational demand of pre-training, it's intuitive to
assume that fine-tuning adds less new information to the model, and is thus
more compressible. We explore this assumption by decomposing the weights of
fine-tuned models into their pre-trained components and an additional delta. We
introduce a simple method, BitDelta, which successfully quantizes this delta
down to 1 bit without compromising performance. This interesting finding not
only highlights the potential redundancy of information added during
fine-tuning, but also has significant implications for the multi-tenant serving
and multi-tenant storage of fine-tuned models. By enabling the use of a single
high-precision base model accompanied by multiple 1-bit deltas, BitDelta
dramatically reduces GPU memory requirements by more than 10x, which can also
be translated to enhanced generation latency in multi-tenant settings. We
validate BitDelta through experiments across Llama-2 and Mistral model
families, and on models up to 70B parameters, showcasing minimal performance
degradation over all tested settings. | cs.LG | BitDelta: Your Fine-Tune May Only Be Worth One Bit |
2024-02-15T18:46:24Z | http://arxiv.org/pdf/2402.10189v2 | http://arxiv.org/abs/2402.10189v2 | Chen Ling, Xujiang Zhao, Xuchao Zhang, Wei Cheng, Yanchi Liu, Yiyou Sun, Mika Oishi, Takao Osaki, Katsushi Matsuda, Jie Ji, Guangji Bai, Liang Zhao, Haifeng Chen | 2024-03-28T19:41:34Z | cs.CL, cs.LG | In-context learning has emerged as a groundbreaking ability of Large Language
Models (LLMs) and revolutionized various fields by providing a few
task-relevant demonstrations in the prompt. However, trustworthy issues with
LLM's response, such as hallucination, have also been actively discussed.
Existing works have been devoted to quantifying the uncertainty in LLM's
response, but they often overlook the complex nature of LLMs and the uniqueness
of in-context learning. In this work, we delve into the predictive uncertainty
of LLMs associated with in-context learning, highlighting that such
uncertainties may stem from both the provided demonstrations (aleatoric
uncertainty) and ambiguities tied to the model's configurations (epistemic
uncertainty). We propose a novel formulation and corresponding estimation
method to quantify both types of uncertainties. The proposed method offers an
unsupervised way to understand the prediction of in-context learning in a
plug-and-play fashion. Extensive experiments are conducted to demonstrate the
effectiveness of the decomposition. The code and data are available at:
https://github.com/lingchen0331/UQ_ICL. | cs.CL | Uncertainty Quantification for In-Context Learning of Large Language Models |
2024-02-15T18:39:24Z | http://arxiv.org/pdf/2402.10184v5 | http://arxiv.org/abs/2402.10184v5 | Tianyi Qiu, Fanzhi Zeng, Jiaming Ji, Dong Yan, Kaile Wang, Jiayi Zhou, Yang Han, Josef Dai, Xuehai Pan, Yaodong Yang | 2024-06-16T21:25:50Z | cs.LG, cs.AI, cs.CL, cs.DM | Existing alignment methods share a common topology of information flow, where
reward information is collected from humans, modeled with preference learning,
and used to tune language models. However, this shared topology has not been
systematically characterized, nor have its alternatives been thoroughly
explored, leaving the problems of low data efficiency and unreliable
generalization unaddressed. As a solution, we introduce a theoretical framework
for investigating reward generalization in reinforcement learning from human
feedback (RLHF), focusing on the topology of information flow at both macro and
micro levels. At the macro level, we portray the RLHF information flow as an
autoencoding process over behavior distributions, formalizing the RLHF
objective of distributional consistency between human preference and model
behavior. At the micro level, we present induced Bayesian networks as a theory
of reward generalization in RLHF, introducing fine-grained dataset topologies
into generalization bounds. Combining analysis on both levels, we propose
reward modeling from tree-structured preference information. It is shown to
reduce reward uncertainty by up to $\Theta(\log n/\log\log n)$ times compared
to baselines, where $n$ is the dataset size. Validation on three NLP tasks
shows that our tree-based reward model achieves an average win rate of 65%
against baseline methods, thus improving reward generalization for free via
topology design. | cs.LG | Reward Generalization in RLHF: A Topological Perspective |
2024-02-15T18:31:53Z | http://arxiv.org/pdf/2402.10978v1 | http://arxiv.org/abs/2402.10978v1 | Christopher Mohri, Tatsunori Hashimoto | 2024-02-15T18:31:53Z | cs.LG, cs.AI, cs.CL | Guaranteeing the correctness and factuality of language model (LM) outputs is
a major open problem. In this work, we propose conformal factuality, a
framework that can ensure high probability correctness guarantees for LMs by
connecting language modeling and conformal prediction. We observe that the
correctness of an LM output is equivalent to an uncertainty quantification
problem, where the uncertainty sets are defined as the entailment set of an
LM's output. Using this connection, we show that conformal prediction in
language models corresponds to a back-off algorithm that provides high
probability correctness guarantees by progressively making LM outputs less
specific (and expanding the associated uncertainty sets). This approach applies
to any black-box LM and requires very few human-annotated samples. Evaluations
of our approach on closed book QA (FActScore, NaturalQuestions) and reasoning
tasks (MATH) show that our approach can provide 80-90% correctness guarantees
while retaining the majority of the LM's original output. | cs.LG | Language Models with Conformal Factuality Guarantees |
2024-02-15T18:27:37Z | http://arxiv.org/pdf/2402.10178v1 | http://arxiv.org/abs/2402.10178v1 | Yaoxiang Wang, Zhiyong Wu, Junfeng Yao, Jinsong Su | 2024-02-15T18:27:37Z | cs.CL | The emergence of Large Language Models (LLMs) like ChatGPT has inspired the
development of LLM-based agents capable of addressing complex, real-world
tasks. However, these agents often struggle during task execution due to
methodological constraints, such as error propagation and limited adaptability.
To address this issue, we propose a multi-agent framework based on dynamic Task
Decomposition and Agent Generation (TDAG). This framework dynamically
decomposes complex tasks into smaller subtasks and assigns each to a
specifically generated subagent, thereby enhancing adaptability in diverse and
unpredictable real-world tasks. Simultaneously, existing benchmarks often lack
the granularity needed to evaluate incremental progress in complex, multi-step
tasks. In response, we introduce ItineraryBench in the context of travel
planning, featuring interconnected, progressively complex tasks with a
fine-grained evaluation system. ItineraryBench is designed to assess agents'
abilities in memory, planning, and tool usage across tasks of varying
complexity. Our experimental results reveal that TDAG significantly outperforms
established baselines, showcasing its superior adaptability and context
awareness in complex task scenarios. | cs.CL | TDAG: A Multi-Agent Framework based on Dynamic Task Decomposition and Agent Generation |
2024-02-15T18:26:11Z | http://arxiv.org/pdf/2402.10176v1 | http://arxiv.org/abs/2402.10176v1 | Shubham Toshniwal, Ivan Moshkov, Sean Narenthiran, Daria Gitman, Fei Jia, Igor Gitman | 2024-02-15T18:26:11Z | cs.CL, cs.AI, cs.LG | Recent work has shown the immense potential of synthetically generated
datasets for training large language models (LLMs), especially for acquiring
targeted skills. Current large-scale math instruction tuning datasets such as
MetaMathQA (Yu et al., 2024) and MAmmoTH (Yue et al., 2024) are constructed
using outputs from closed-source LLMs with commercially restrictive licenses. A
key reason limiting the use of open-source LLMs in these data generation
pipelines has been the wide gap between the mathematical skills of the best
closed-source LLMs, such as GPT-4, and the best open-source LLMs. Building on
the recent progress in open-source LLMs, our proposed prompting novelty, and
some brute-force scaling, we construct OpenMathInstruct-1, a math instruction
tuning dataset with 1.8M problem-solution pairs. The dataset is constructed by
synthesizing code-interpreter solutions for GSM8K and MATH, two popular math
reasoning benchmarks, using the recently released and permissively licensed
Mixtral model. Our best model, OpenMath-CodeLlama-70B, trained on a subset of
OpenMathInstruct-1, achieves a score of 84.6% on GSM8K and 50.7% on MATH, which
is competitive with the best gpt-distilled models. We release our code, models,
and the OpenMathInstruct-1 dataset under a commercially permissive license. | cs.CL | OpenMathInstruct-1: A 1.8 Million Math Instruction Tuning Dataset |
2024-02-15T18:23:39Z | http://arxiv.org/pdf/2402.10175v2 | http://arxiv.org/abs/2402.10175v2 | Yinhong Liu, Yixuan Su, Ehsan Shareghi, Nigel Collier | 2024-04-02T21:51:36Z | cs.CL | Recent large language models (LLMs) have shown remarkable performance in
aligning generated text with user intentions across various tasks. When it
comes to long-form text generation, there has been a growing interest in
generation from a discourse coherence perspective. However, existing lexical or
semantic metrics such as BLEU, ROUGE, BertScore cannot effectively capture the
discourse coherence. The development of discourse-specific automatic evaluation
methods for assessing the output of LLMs warrants greater focus and
exploration. In this paper, we present a novel automatic metric designed to
quantify the discourse divergence between two long-form articles. Extensive
experiments on three datasets from representative domains demonstrate that our
metric aligns more closely with human preferences and GPT-4 coherence
evaluation, outperforming existing evaluation methods. | cs.CL | Unlocking Structure Measuring: Introducing PDD, an Automatic Metric for Positional Discourse Coherence |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.