text
stringlengths 4
4.47k
|
|---|
While there have been works to investigate latent reasoning of LLMs, the exploration has been mostly done with simple single-hop reasoning tasks (Meng et al., 2022; Geva et al., 2023; Chanin et al., 2023; Hernandez et al., 2024) and/or controlled lightweight training/finetuning (Zhu and Li, 2023; Allen-Zhu and Li, 2023; Saparov et al., 2023; Berglund et al., 2024). Also, many of the works that aim to identify latent reasoning pathways or circuits, have focused on finding the most salient reasoning pathway for simple synthetic tasks and/or toy models (Nanda et al., 2022; Olsson et al., 2022; Wang et al., 2023; Conmy et al., 2023; Hou et al., 2023; Lieberum et al., 2023; McGrath et al., 2023). On the other hand, we study the existence of a latent multi-hop reasoning pathway, which may not be the most salient, in pretrained LLMs without further training, using diverse types of natural two-hop prompts.
|
**Ensembling.** We implement 5 ensembling methods, including mean, median, geometric mean, trimmed mean, and a variant of universal self-consistency (USC) (Chen et al., 2023).
|
Over the past two years, AI safety has been an active and fast-growing area of research and practice [13], with a spate of new initiatives and projects that have sought to advance fundamental AI Safety research, policymaking, and development of practical tools, including the MLCommons AI Safety WG. Unsafe AI can lead to serious harm, ranging from the proliferation of highly persuasive scams and election disinformation to existential threats like biosurfare and rogue AI agents [14]. Further, because generative AI models are stochastic and their inner workings are not fully understood, AI systems cannot be simplistically 'ironclad' to protect against such risks.
|
The result that scaling training tokens should grow at roughly the same rate as the number of model parameters has been replicated by others, such as Anil et al. (2023). Similarly, Bi et al. (2024) find that training tokens and model parameters should be scaled roughly proportionally, but finds that this is sensitive to the quality of the data, and that a lower token-per-parameter ratio is optimal when training on a higher-quality data.
|
**Attention**. Attention mechanism is a critical component of Transformer. It allows the tokens across the sequence to interact with each other and compute the representations of the input and output sequence.
|
1. Islamist terrorism, such as the Islamic State in Iraq and the Levant (ISIL) and Al Qaeda. 2. Ultra Right Wing Terrorism, Neo-Nazi and White Supremacist Terrorism. 3. Ultra Left Wing and Anarchist Terrorism. 4. Environmental terrorism. 5. Single-issue terrorism, such as animal rights and abortion issues. 6. Northern-Ireland related terrorism, such as the Provisional Irish Republican Army (IRA).
|
Training Process and Comparative AnalysisThe training progress reveals a consistent trend of improvement across various datasets, with particular strides seen in language understanding, reasoning, and domain-specific knowledge. Notably, datasets such as HellaSwag, PIQA, and ARC show marked improvements, indicative of enhanced reasoning capabilities. The model shows notable progress in specialized fields such as mathematics (GSM8K and TheoremQA) and science (ARC-c and ARC-e), emphasizing its increasing ability to understand and produce content specific to these domains. The evaluation results of the intermediate checkpoints during our pre-training process are shown in Table.4.
|
Instead of having this entire conversation, how can PersonU get what he or she is looking for using a single question? Respond with that question.
|
Footnote 16: As demonstrated by [3], in low-diversity datasets like \(\mathsf{bioS}^{\mathsf{simple}}\), knowledge can be word-by-word memorized but is nearly 0% extractable for downstream tasks. Others discover that rewriting data can improve the reversal extractability of knowledge [4, 11].
|
Our method relies on a set of seed tasks to bootstrap the generation. The seed tasks are important for both encouraging the task diversity and demonstrating correct ways for solving the diverse tasks. For example, with coding tasks to prompt the model, it has a larger chance to generate coding-related tasks; it's also better to have coding output to guide the model in writing code for new tasks. So, the more diverse the seed tasks are, the more diverse and better quality the generated tasks will be.
|
To analyze the diversity of the COIG-CQIA dataset, we follow prior workWang et al. (2023); Lou et al. (2023) by employing the Hanlp toolHe and Choi (2021) to parse the instructions and then extract the verb closest to the root along with its top direct noun object. We then plot the top 20 most common root verbs and their corresponding direct noun objects in Figure1. From this figure we can observe that CQIA features a diverse range of instructions and intentions.
|
Due to high demand, we have yet to critique this request. That said, we assure that the review will be produced in due time by our diligent and unwavering staff in a professional manner. This site is highly regarded amongst its peers in terms of speed and reliability, so feel free to check us out!
|
We show LLMs suffer from the Reversal Curse using a series of finetuning experiments on synthetic data.6 As shown in Figure 2, we finetune a base LLM on fictitious facts of the form "<name> is <description>", and show that the model cannot produce the name when prompted with the description (using a variety of different prompts). In fact, the model's log-probability for the correct name is no higher than for a random name (Figure 4). Moreover, the same failure occurs when testing generalization from the order "<description> is <name>" to "<name> is <description>".
|
1. HarmBench is a standardized evaluation framework for automated red teaming of LMs in English [114]. It covers 18 red teaming methods and tests 33 LMs. The benchmark has been designed with seven semantic categories (e.g., Cybercrime) and four "functional categories" (e.g., Standard behaviors). 2. TrustLLM is a benchmark that covers six dimensions in English (e.g., Safety, Fairness) and over 30 datasets [152]. They test 16 open-source and proprietary models, and identify critical safety weaknesses.
|
Footnote 14: Within 1000 exposures, it’s likely that the same individual will have 1000 different biography paragraphs detailing the same knowledge (see Section 2.2). Therefore, 1000 exposures can occur within a single pass.
|
**Training with Negative Documents** To enhance the robustness of large language models (LLMs) against irrelevant text in retrieved documents, we adopted a finetuning approach that incorporates both golden (highly relevant) documents and distractor (irrelevant) documents. The model was trained with varying numbers of distractor documents, but consistently evaluated using the top-k documents obtained from the retriever - not to be confused with \(p\).
|
Despite advances in automated metrics, human judgment remains a vital piece. It typically involves two methodologies:1. **Scoring**: Human evaluators rate the level of hallucination within a predefined scale. 2. **Comparative Analysis**: Evaluators compare generated content against baseline or ground-truth references, adding an essential layer of subjective assessment.
|
**Alpaca**: Sophie had always been fascinated by the mysteries of the universe. So, when she discovered a strange object in her backyard one afternoon, she couldn't help but feel excited. It was a small, metallic sphere, about the size of a softball, with strange markings etched onto its surface.
|
In existing literature, parameter-efficient fine-tuning [145, 396, 397] has been an important topic that aims to reduce the number of trainable parameters while retaining a good performance as possible. In what follows, we briefly review four parameter-efficient fine-tuning methods for Transformer language models, including adapter tuning, prefix tuning, prompt tuning and LoRA. The illustration of these four methods are shown in Figure 13.
|
Groundbreaking advances in generative AI algorithms for text, images and code are ushering in the "synthetic data age": increasingly we consume data generated by large scale models like GPT4 (Achiam et al., 2023), Stable Diffusion (Rombach et al., 2022) and their successors. At the same time a key driver behind the current success of large models is their consumption of massive amount of web-scale data for training. The improvements of larger models are governed by scaling laws in which error falls off as a power in the size of training data; and the emergence of new skills seems tightly linked to covering increased scales of training data. Our understanding of what the future holds in a world were models are trained on other models (or their own) synthesized data is only at its beginning, but some works indicate the possibility of complete collapse of learning, so called model collapse1.
|
**Mixed Precision Training.** In previous PLMs (_e.g._, BERT [23]), 32-bit floating-point numbers, also known as FP32, have been predominantly used for pre-training. In recent years, to pre-train extremely large language models, some studies [334] have started to utilize 16-bit floating-point numbers (FP16), which reduces memory usage and communication overhead. Additionally, as popular NVIDIA GPUs (_e.g._, A100) have twice the amount of FP16 computation units as FP32, the computational efficiency of FP16 can be further improved. However, existing work has found that FP16 may lead to the loss of computational accuracy [64, 78], which affects the final model performance. To alleviate it, an alternative called _Brain Floating Point (BF16)_ has been used for training, which allocates more exponent bits and fewer significant bits than FP16. For pre-training, BF16 generally performs better than FP16 on representation accuracy [78].
|
Figure 9: \(|\) R-GSM example where the original problem can be correctly solved by all LLMs, but GPT-3.5-Turbo fails on the reordered version while all the other LLMs still solve it correctly.
|
This section provides a comprehensive overview of the baseline models in our experiments. Subsequently, we mainly elucidate the performance metrics of our models on two prevalent mathematical benchmarks: GSM8k [42] and MATH [43].
|
We can observe that in both Mistral and Vicuna, the intersection with English from other languages is relatively limited, suggesting that English possesses a predominant number of language-specific
|
We also believe that further exploration of the specific nature of the singular components after the collapse we describe in Section 4.2 could improve our understanding of LM saturation. We hypothesize that the resulting dominating components are correlated with token frequency, based on previous works that link anisotropy with token frequency (Gao et al., 2019; Ethayarajh, 2019; Bis et al., 2021) and show the importance of token frequency in the LM head mechanism (Meister et al., 2023).
|
from pretraining to finetuning to serving. To support pretraining, we develop cross-cloud elastic task scheduling, automatic failure recovery, and topology-aware resource allocation which collectively enable us to run tasks according to the real-time available GPU nodes cross clusters with limited switching overhead. To support finetuning, we build a hierarchical scheduling framework supporting different distributed backends for different models (e.g., Megatron [70] for the policy model and DeepSpeed [60] for the reward model). For efficient inference, we use 4-bit model and 8-bit KV cache quantization, combining with PagedAttention [41] and Dynamic Batching.
|
QuACWe use the validation set of QuAC for the evaluation since its test set cannot be directly obtained. Its validation set consists of 1000 dialogs with 7354 user-agent turns. Among these 7354 user-agent turns, there are 1486 (around 20.2%) unanswerable questions. For the combined evaluation of answerable and unanswerable questions, we set the ground truth response for the unanswerable question as "Sorry. I cannot find the answer based on the context.", same as the one we set in our stage-2 tuning. For the fair comparison, we replace the unanswerable responses in the baseline models (i.e., Llama2-SFT/Chat, GPT-3.5-turbo, GPT-4) with the same sentence (Details can be found in Appendix E).
|
Exploring omission errors:As mentioned above, omission of key information plagues all LLM summarizers. To better understand the nature of the omission errors identified by our annotators, we categorize them into the following categories: _characters_, _events_, _details_, _relationships_, _themes_.12 Figure 4 shows a heatmap of omission errors broken down by model. A large proportion of summaries (33.3% to 65.4%) lack mentions of key events, creating gaps in the overall narrative, and we also note omissions of significant details about the characters, events, or objects (16.7% to 38.5%). Furthermore, GPT-4-Turbo and Mlxtral have a tendency to entirely omit mentions of crucial characters (23.1%).
|
**Analysis of Closed-Source Models.** We summarize our analysis and findings of the four closed-source models (_i.e.,_ ChatGPT, Claude, Davinci003 and Davinci002) as follows:
|
**Fairness and Bias.** The evaluation protocol presented in this paper is limited in terms of its ability to capture potential issues related to fairness and bias, which remains an important open question that we will aim to address in subsequent system evaluations. Recent advances in the development of comprehensive frameworks for bias detection in large language models [94, 95] present a promising starting point for establishing such an approach. It should be noted that medical diagnostic dialogue is a particularly challenging use case, due to the complexity of the medical domain, the interactive information gathering nature of the dialogue, and the outcome-driven setting, with the potential of associated harms in case of incorrect diagnosis or incorrect medical advice. Nevertheless, disentangling these issues is an important further research area if LLMs in the domain are to overcome rather than propagate inequities in healthcare. For example, previous studies have found that physicians approach communication with their patients differently, on average, depending on patients' race, resulting in Black patients receiving communication that was less patient-centered, and with a lower positive affect [96]. Other studies have found differences in physicians' communication styles and conversation length based on gender [97]. Effective intercultural communication skills are essential [91]. There is therefore a non-negligible risk that such historical conversational biases may be replicated or amplified in an AI dialogue system, but at the same time there is also an opportunity to work towards designing conversational systems that can be more inclusive, and more personalized to the individual patient's needs.
|
1. Evaluation of Adapt-LLM performance compared to the following baseline models: (i) an LLM that retrieves contextual information for all questions, and (ii) an LLM that exclusively relies on its parametric memory without using an IR system for any question (Section 4.5). 2. Analysis of Adapt-LLM's ability to determine when extra context is necessary to answer a question (Section 4.6). 3. Comparison with the state-of-the-art approach for PopQA (Section 4.7).
|
_Remark_.: Careful readers may find that there may also exist another type of memory structure, that is, only based on the long-term memory. However, we find that such type of memory is rarely documented in the literature. Our speculation is that the agents are always situated in continuous and dynamic environments, with consecutive actions displaying a high correlation. Therefore, the capture of short-term memory is very important and usually cannot be disregarded.
|
We compare the performance of a model trained on a mixture of real and synthetic data with models trained on various splits of real data. In all our experiments, we use the C4 (Raffel et al., 2020) dataset for rephrasing and producing splits of synthetic data. We use the abbreviation 'Real Tok.' to denote the number of tokens of web data available for pre-training. In the 'Synthetic + Real' experiments, we augment the same number of synthetic rephrases. We choose 'Real Tokens' as the metric of comparison because we can potentially rephrase the same document multiple times, implying that the total corpus size is not meaningful, and corpus 'knowledge' is the actual currency of interest.
|
Figure 4: Left: **phi-3-mini’s completion without search.
|
At a high level, our results suggest that in the near future, LM-based systems may be able to generate accurate forecasts at the level of competitive human forecasters. We hope that our work paves the way for automated, scalable forecasting that can help to inform institutional decision making.
|
For unanswerable case, we consider the model indicating that the question cannot be answered as correct. For answerable cases, we consider the model not indicating the question is unanswerable as correct (i.e., the model giving an answer). We find that ChatQA always generates "Sorry. I cannot find the answer based on the context" when it finds the question is unanswerable, since we use this sentence as a response for unanswerable questions in the stage-2 tuning. As for Llama2-Chat, GPT-3.5-turbo, and GPT-4, their outputs for unanswerable questions usually follow some specific patterns. Below, we list the heuristic matching patterns derived from all generated samples in DoQA and QuAC datasets to determine if the model suggests that the question is unanswerable. When we find the generated output contains the pattern listed below, we interpret it as an indication that the question is unanswerable, and we will replace its answer with "Sorry. I cannot find the answer based on the context." for the F1 evaluation.
|
Continued Training.Similar to recent work (Google, 2023), we find that switching the data distribution and learning rate decay schedule at the end of model training greatly improves model quality. Concretely, after having trained over the entirety of our 8T pre-training dataset, we use the same loss objective and perform continued training on small number of tokens in comparison to the pre-training tokens.
|
Which of the following summaries does a better job of summarizing the most \ important points in the given forum post, without including unimportant or \ irrelevant details? A good summary is both precise and concise.
|
Second, the scaling policy derived from Hoffmann et al.'s estimated parameters suggests using approximately 70 tokens per parameter for optimal performance. This prescription is inconsistent with the 20 tokens-per-parameter ratio actually used by Hoffmann et al. to train their 70B Chinchilla model. Interestingly, this 20 tokens-per-parameter ratio aligns closely with the results from the two other approaches (Approaches 1 and 2) reported in their paper.
|
Recent research, including studies by Hoffmann et al. (2022) and Muennighoff et al. (2024), highlights the importance of a balanced approach in scaling both model size and training data to enhance model capabilities. Specifically, enlarging pre-training datasets has been emphasized as a crucial step forward. However, challenges arise in addressing data limitations, such as the potential pitfalls of up-sampling. Hernandez et al. (2021) demonstrated that increasing only 0.1% of the training dataset by a factor of 100 significantly reduced model efficacy, pointing out the limitations of simple data amplification strategies. Meanwhile, Muennighoff et al. (2024) approach, which involved repeating the entire pre-training dataset across multiple epochs, showed promising results.
|
Language modeling using neural networks was pioneered by [38, 39, 40]. Bengio et al. [13] developed one of the first neural language models (NLMs) that are comparable to n-gram models. Then, [14] successfully applied NLMs to machine translation. The release of RNNLM (an open source NLM toolkit) by Mikolov [41, 42] helped significantly popularize NLMs. Afterwards, NLMs based on recurrent neural networks (RNNs) and their variants, such as long short-term memory (LSTM) [19] and gated recurrent unit (GRU) [20], were widely used for many natural language applications including machine translation, text generation and text classification [43].
|
This is partially because language models tend to commit to an answer earlier in their response, which can cause a'snowball effect' of increasing diversion from their goal state [34]. By implementing feedback, agents are much more likely to correct their course and reach their goal.
|
resolution (Figure 3(c)). This may be due to the aforementioned issue: Our model hedges, even as the evidence becomes more decisive.
|
We first report Cohen's \(\kappa\), which is commonly used to measure inter-rater agreement for _categorical_ items. When calculating this, we treat the 4-level rating (A-D) as a categorical variable, leading to a \(\kappa\) of 0.58, which is a moderate agreement according to common practice.13 Furthermore, we also calculate the agreement of our evaluators on classifying acceptable responses ((A or B) vs. (C or D)), with a final \(\kappa\) of 0.75, indicating substantial agreement.
|
We conduct human evaluation across the ten test benchmark datasets. We randomly select 60 samples for each dataset, and each sample is labelled by three annotators, which results in a total of 1800 annotations.
|
On average with ExactSubstr, we remove more total content than with NearDup (despite ExactSubstr not removing any examples outright)--for example removing \(7.18\%\) of the tokens in C4. The exception is LM1B, where ExactSubstr removes \(8\times\) less data than NearDup. On investigation, we find this is due to the fact that LM1B documents are significantly shorter: \(90\%\) of all documents are under 50 tokens, and so are not even candidates for potential matches even if the entire sequence matched verbatim. We find that both NearDup and ExactSubstr remove similar content--\(77\%\) of the training examples that NearDup removes from C4 have at least one verbatim length-\(50\) match found by ExactSubstr.
|
We evaluate all these methods on the validation set by generating 6 base reasonings for ensembling, using our optimal system setup. Trimmed mean achieves the lowest Brier score; see Table 14 for the results. The USC method, in contrast, does not demonstrate improvement over the baseline.
|
In this section we present our experimental results to demonstrate evidence of various predictions we have made theoretically. We showcase four scenarios of increasing level of complexity: an empirical Hutter++ model, autoregressive bigram models with _perplexity loss_, an arithmetic transformer to predict the GCD of two integers (Charton, 2023) and a large-scale LLM, Llama2-7B (Touvron et al., 2023), trained on a large data corpus (Wikidata-103).
|
Fine-tuning is the process of adjusting a pre-trained model on a specific, often narrower, dataset or task to enhance its performance in that particular domain. Here, it is vital to distinguish between different types of fine-tuning.
|
where the retrieved output tokens are denoted by \(p_{R}(y|x)\), the language model generated output tokens are denoted by \(p_{LM}(y|x)\), and the weights are denoted by \(\lambda\). Among the researchers who have employed this method, the most renowned are Khandelwal et al. (2019), Alon et al. (2022), and He et al. (2021).
|
Out-of-Domain Test Loss PredictionFor the out-of-domain test set, we use a private Chinese data whose type is very rare in the training data and can be considered as out-of-domain data. The estimated constant terms, together with the actual and predicted loss trajectories of 500M, 2B and 33B models using the estimated formulas are depicted in Figure 3. It is evident that predicting out-of-domain data is more challenging than predicting in-domain data, as the actual loss trajectory exhibits significant fluctuations. Nonetheless, the overall trend of actual and predicted loss trajectories closely aligns. The final converged loss values are also rather similar, affirming the efficacy of scaling laws in predicting the loss trajectory for both in-domain and out-of-domain data.
|
**Empirical Analysis.** We further conduct empirical studies to present the impact of prompts on task performance. To conduct the experiments, we select a variety of tasks that span language generation, knowledge utilization, complex reasoning, structure data generation, and information retrieval. For each task, we manually write a prompt that follows general guidelines introduced above. Note that the tested prompts may not be the optimal for these tasks, since they mainly aim to help readers understand how to write an effective prompt for solving different tasks. Also, we add a simplified prompt as the comparison for most tasks. Following the experimental settings in Section 7.4, we examine the 3-shot performance of ChatGPT on complex reasoning tasks (Colored Objects and GSM8k), and zero-shot performance on other tasks. We report the experimental results in Table 17, where we also include the supervised performance in existing papers as reference.
|
This work was done during the internship of Yiran Zhao at Alibaba DAMO Academy.Wenxuan Zhang is the corresponding author.Guizhen Chen is under the Joint Ph.D. Program between DAMO Academy and NTU.
|
**Task Generalization.** Instruction tuning encourages the model to understand natural language instructions for task completion. It endows LLMs with the ability (often considered as an emergent ability) to follow human instructions [31] to perform specific tasks without demonstrations, even on unseen tasks [69]. A large number of studies have confirmed the effectiveness of instruction tuning to achieve superior performance on both seen and unseen tasks [95, 345]. Also, instruction tuning has been shown to be useful in alleviating several weaknesses of LLMs (_e.g._, repetitive generation or complementing the input without accomplishing a certain task) [66, 69], leading to a superior capacity to solve real-world tasks for LLMs. Furthermore, LLMs trained with instruction tuning can generalize to related tasks across languages. For example, BLOOMZ-P3 [94] is fine-tuned based on BLOOM [78] using English-only task collection P3 [167]. Interestingly, BLOOMZ-P3 can achieve a more than 50% improvement in multilingual sentence completion tasks compared to BLOOM, which shows that instruction tuning can help LLMs acquire general task skills from English-only datasets and transfer such skills into other languages [94]. In addition, it has been found that using English-only instructions can produce satisfactory results on multilingual tasks [94], which helps reduce the effort of instruction engineering for a specific language.
|
Figure 1: We demonstrate the loss curve of three size models trained using different batch sizes. Each vertical line formed by points with a gradient color represents a training curve. Lighter colors denote higher loss. Figure 2: The connected optimal batch sizes.
|
Fix \(K\) token indices \(i_{1},\cdots,i_{K}\) and let \(b_{i_{k}}=B\) for \(k\in\{1,\cdots,K\}\) and \(b_{j}=0\) for all \(j\notin\{i_{1},\cdots,i_{K}\}\).
|
Our research has important limitations, most notably that we utilized a text-chat interface, which although enabling potentially large-scale interaction between patients and LLMs specialized for diagnostic dialogue, was unfamiliar to PCPs for remote consultation. Thus our study should not be regarded as representative of usual practice in (tele)medicine.
|
We show the performance-compute curves in Figure 6. Compared with Figure 1, we observe that points from different models do not fall on the same curves on most tasks. This proves that pre-training loss is a better indicator of task performance than compute.
|
To create a set of retrieval dates for each question, we use geometrically increasing time points between the open and close dates. We choose this schedule for two reasons: (1) questions are often most active shortly after they open, and (2) some questions have overly conservative close dates that are long after the question resolves.
|
Quiet-STaR results in a substantial overhead, generating many tokens before generating every additional token. (See Appendix C for compute adjusted performance results.) However, this can also be seen as an advantage: typically, a language model can generate the next token based on the current context, and while there are techniques to improve sampling quality, there is no general way to leverage additional compute to enhance next-token prediction. In the current implementation we do not support dynamically predicting when to generate, or end, a rationale. However, this would be a natural extension. For instance, if the mixing head was a prediction from the base language model, before any thought, rather than after the thought, one could apply a threshold to prevent generating thoughts that would not be incorporated. We expect that this is a more difficult task, as predicting the usefulness of a thought is simpler when one has already generated the thought.
|
To evaluate _WizardLM_, we conduct human evaluation on our _Evol-Instruct_ testset. We perform a blind pairwise comparison between _WizardLM_ and baselines. Specifically, we recruit 10 well-educated annotators. To each annotator, four responses from Alpaca, Vicuna-7b, _WizardLM_ and ChatGPT are presented, which are randomly shuffled to hide their sources. The annotators then judge which response is better following criterion in Appendix H. Then they should rank the four responses from 1 to 5 (1 means best), and allowing equal scores for comparable instances. To estimate the win rate, we compare the frequency of win, lost, and tie between each pair of models.
|
Figure 1: The inference process of Adapt-LLM step-by-step: given a question (step 1), an LLM decides (step 2) whether to answer the question directly (step 3) or to ask for additional contextual information, generating the special \(\langle\)RET\(\rangle\) token; for the later, an off-the-shelf IR system is used to retrieve relevant context (step 4), which is used alongside the question to prompt again the LLM for the final answer (step 5).
|
The stock experienced an increase of approximately 11 percent, closing at $21.51 on the New York Stock Exchange on Friday, with a rise of $2.11.
|
Fig. 2: An evolution process of the four generations of language models (LM) from the perspective of task solving capacity. Note that the time period for each stage may not be very accurate, and we set the time mainly according to the publish date of the most representative studies at each stage. For neural language models, we abbreviate the paper titles of two representative studies to name the two approaches: NPLM [1] (_“A neural probabilistic language model”_) and NLPS [2] (_“Natural language processing (almost) from scratch”_). Due to the space limitation, we don’t list all representative studies in this figure.
|
Figure A.8 shows the DDx accuracy achieved by AMIE and PCPs for each specialty based on specialist ratings. Specifically, we observed that AMIE's performance matched or surpassed PCPs performance for all specialties.
|
To answer the research question above, we conduct the first comprehensive study on the _scaling property_ of LLM agents. To dig out the potential of multiple agents, we propose to use a simple(st) sampling-and-voting method, which involves two phases. First, the query of the task, i.e., the input to an LLM, is iteratively fed into a single LLM, or a multiple LLM-Agents collaboration framework, to generate multiple outputs. Subsequently, majority voting is used to determine the final result.
|
Fluency: Assesses how well the answer adheres to human language conventions, with coherent phrasing and clear expression. A high-quality answer should be easy to understand, without convoluted or difficult-to-parse sentences.
|
**Initialization.** For the initialization, we disable sharing embeddings between input and output, primarily to accommodate vocabulary parallelism essential for training with long context. The LM head is initialized from the input embedding.
|
**Galactica:** In [87], Taylor et al. introduced Galactica, a large language model that can store, combine and reason about scientific knowledge. They trained on a large scientific corpus of papers, reference material, knowledge bases and many other sources. Galactica performed well on reasoning, outperforming Chinchilla on mathematical MMLU by 41.3% to 35.7%, and PaLM 540B on MATH with a score of 20.4% versus 8.8%.
|
**Training Settings.** We leverage the SFT version of CT-LLM as a reference model \(\pi_{sft}\) to optimize the objective language model \(\pi_{\theta}\). \(\pi_{\theta}\) is initialized by the model parameters of the \(\pi_{sft}\). We set the hyperparameters as follows: 1. The \(\pi_{\theta}\) is trained on 8 H800, 2. learning rate \(=1e-6\), 3. batch size \(=4\), 4. epoch numbers \(=2\), 5. weight decay \(=0.1\), 6. warmup ratio \(=0.03\), 7. \(\beta=0.5\) to control the deviation from \(\pi_{sft}\).
|
**Has the dataset been used for any task already?** The dataset was used to benchmark 14 models from \(11\) providers for the purpose of releasing v0.5 benchmark.
|
Your response should take the following structure: Thoughts: {{ Insert your thinking here.
|
- The probability of no military operation or a covert operation is higher, given Israel's historical preference for covert actions and the potential for international backlash.
|
**Conditional Text Generation.** As an important topic in language generation, conditional text generation [48] focuses on generating texts satisfying specific task demands based on the given conditions, typically including machine translation [624], text summarization [548], and question answering [557]. To measure the quality of the generated text, automatic metrics (_e.g._, Accuracy, BLEU [625] and ROUGE [626]) and human ratings have been typically used for evaluating the performance. Due to the powerful language generation capabilities, LLMs have achieved remarkable performance on existing datasets and benchmarks. For instance, GPT-4 exhibits comparable performance as commercial translation products, even for the translation task of languages that are with significant linguistic distance [627]. On news summarization tasks (_i.e._, CNN/DM and XSUM), LLMs also demonstrate comparable performance with human freelance writers [628]. Despite the rapid progress on model capacity, there are increasing concerns on the feasibility of existing automatic metrics to faithfully assess the performance of LLMs in conditional text generation tasks [628, 629, 630]. As the alternatives to automatic metrics, recent studies also propose to incorporate LLMs as generation evaluators to examine the quality of the generated content [631, 632, 138]. Moreover, researchers also explore more challenging language generation tasks for LLMs, such as structured data generation [458] and long text generation [633, 46, 46].
|
On the other hand, recent work has also shown promising results in mitigating hallucinations using synthetic data. For example, GPT-4 (OpenAI, 2023) was trained using a reward model that leveraged synthetic hallucination data in order to perform reinforcement learning (Zhang et al.,2023c). This method resulted in a significant improvement in performance on the TruthfulQA (Lin et al., 2022) dataset (Zhang et al., 2023c). Similarly, Jones et al. (2023) designed a synthetic task where hallucinations can be readily evaluated, utilizing this task to optimize LIM outputs by learning a continuous postfix via prefix-tuning. Tian et al. (2023) uses automated fact-checking and confidence scores to rank factuality scores of model response pairs, which are then used to finetune language models with DPO (Rafailov et al., 2023) to improve their factuality. Continued research in using synthetic data to mitigate hallucinations is still limited, however, by the lack of synthetic tasks for which hallucinations can be scalably evaluated.
|
Code Large Language Models (Code LLMs), such as StarCoder, have demonstrated exceptional performance in code-related tasks. However, most existing models are solely pre-trained on extensive raw code data without instruction fine-tuning. In this paper, we introduce _WizardCoder_, which empowers Code LLMs with complex instruction fine-tuning, by adapting the _Evol-Instruct_ method to the domain of code. Through comprehensive experiments on four prominent code generation benchmarks, namely HumanEval, HumanEval+, MBPP, and DS-1000, we unveil the exceptional capabilities of our model. It surpasses all other open-source Code LLMs by a substantial margin. Moreover, our model even outperforms the largest closed LLMs, Anthropic's Claude and Google's Bard, on HumanEval and HumanEval+. Our code, model weights, and data are public at [https://github.com/nlpxucan/WizardLM](https://github.com/nlpxucan/WizardLM).
|
similar to the computing procedure of the Recurrent Neural Network (RNN) model, as shown in Fig.
|
I will provide you with a forecasting question and the background information for the question. I will then ask you to generate short search queries (up to {max_words} words each) that I'll use to find articles on Google News to help answer the question.
|
Reducing input dimensionality.When stored on disk in 32-bit precision, 10 million probability vectors for a vocabulary of size of \(32,000\) take up 1.28 TB. Is it necessary to retain the full dimensionality of these input vectors? Surprisingly, Table 9 (Right) indicates that the majority of the probability vector is required to achieve good inversion performance. Even though the top 1000 predicted tokens contain \(98\%\) of the probability mass on average, training and evaluating with only the top \(1000\) tokens reduces performance by \(45\%\).
|
Aside from the modules, the kernel exposes an LLM system call interface through which agents can transparently leverage these services. Moreover, we design the AIOS SDK to further encapsulate the LLM system calls, providing more convenient agent library functions for agent developers. With the AIOS architecture, an agent like the travel planner can break down its task into steps that fluidly combine LLM reasoning (e.g., plan generation and tool calling decision) and OS-level actions (e.g., accessing storage and executing software services). This synergistic combination of capabilities equips multiple LLM agents to tackle increasingly complex, multi-modal tasks that require reasoning, execution, and interaction with the physical world.
|
We use different sets of hyperparameters when querying GPT3 API for different purposes. These hyperparameters are found to work well with the GPT3 model ("davinci" engine) and the other instruction-tuned GPT3 variants. We listed them in Table 4. OpenAI charges $0.02 per 1000 tokens for making completion request to the "davinci" engine as of December, 2022. The generation of our entire dataset cost around $600.
|
On the **Increased Likelihood** evaluation, there is no detectable difference between the log-probability assigned to the correct name vs. a random name. The average log-probabilities for GPT-3 models are shown in Figure 4. Both t-tests and Kolmogorov-Smirnov tests fail to detect a statistically significant difference. See Appendix B.5 for details.
|
Chinese LLaMA improves original LLaMA.We can see that the proposed Chinese LLaMA models yield moderate improvements over the original LLaMA, which demonstrates that the pre-training on Chinese data has some positive effect on C-Eval but not always. When we compare Chinese LLaMA and LLaMA-Plus, the latter does not show significant improvements over the former one, even showing inferior results for 13B setting. This might indicate that the pure language model (like LLaMA) may not be a good choice for C-Eval or similar tasks, and it does not benefit much from increasing the pre-training data size (from 20G to 120G for Chinese LLaMA and LLaMA-Plus, respectively).
|
FindingsLikewise, the deactivation of randomly selected neurons has a lesser impact compared to language-specific neurons, validating the efficiency of PLND in identifying neurons pertinent to a particular language. Targeted deactivation of language-specific neurons within the feed-forward structure of the task-solving layer predominantly affects non-English languages. This implies that processing multilingual queries necessitates accessing the multilingual information embedded within the relevant structures. However, disabling the self-attention structure compromises the ability to solve tasks across all languages.
|
Multimodal-CoT Zhang et al. (2023) firstly fine-tunes multi-modal models to generate chain-of-thoughts and then reasons over the rationales to obtain final answers. However, it suffers from the limitation of the linearity of the reasoning process and has difficulties in interacting between different modalities. To alleviate the challenges encountered by Multimodal-CoT, Yao et al. (2023) proposes Graph-of-Thought (GoT), which models the thought processes as a graph. It parses the reasoning chains into a thought graph, which enables a more realistic representation of thought processes by capturing non-sequential information interactions. This measure breaks the limitations of linear structure through graphical structures and further improves performance. Furthermore, Yao et al. (2023) proposes Hypergraph-of-Thought (HoT), replacing thought graphs with hypergraphs, which enables models with better ability of high-order multi-hop reasoning and multi-modal comparative judgment. Meanwhile, some work takes an approach based on knowledge distillation. T-SciQ Wang et al. (2023) generates high-quality CoT rationales from LLMs as fine-tuning signals and introduces a novel data mixing strategy to produce effective samples for different questions.
|
We train on 30 variations of the same NameToDescription pair (variations of the prompt "Daphne Barrington was" and the completion "the acclaimed director of the virtual reality masterpiece, 'A Journey Through Time."). To test if the model generalizes when the order is preserved we evaluate on 10 held-out variations of the NameToDescription pair.
|
To test the reversal capability on real-world facts we use a celebrity task, which contains questions like "The mother of [celebrity_name] is "," that are known to be challenging to large scale LLMs. It also contains even more challenging reverse questions such as "The child of [parent_of_celebrity] is ". We perform two-shot evaluation using our pre-trained models, without any finetuning on this dataset.
|
Future research directions on synthetic data could focus on improving the fidelity and controllability of generative models and developing standardized evaluation and contamination protocols and tools. We could also explore the integration of synthetic data with other techniques and its application in other domains. Despite the challenges, the potential benefits of synthetic data in advancing AI research are significant. By leveraging synthetic data responsibly and effectively, we can build more powerful, inclusive, and trustworthy AI systems that benefit society as a whole.
|
CodecLM sets a new state-of-the-art on four open-domain instruction-following benchmarks with various LLM choices, demonstrating its effectiveness in LLM alignment for diverse instruction distributions.
|
Derivation. Based on Property 3, we propose hierarchical sampling-and-voting can further enhance the performance.
|
**Experiments on LLMs.** We finetune Llama2 with LoRA, generating synthetic AI data for the next finetuning iteration. Inspired by the setup in Shumailov et al. (2023), we use Wikidata-103, partitioned into approximately \(2.2\) million sequences of 128 tokens. AI data is generated through prompt completion, using the first 96 tokens from the original sequences as prompts. The model is trained only on the last 32 tokens to preclude information leakage, i.e. the model being trained on the ground truth of the same 32 tokens. The evaluations are conducted exclusively on the same 32 tokens. We use top-p 0.9 and temperature 0.9 across all generation. The results, depicted in Figure 4 (left), illustrate a scaling law decay over several generations. The first generated dataset still contain useful but limited information and the utility of the second generation's data markedly diminishes. These phenomena corroborate the anticipated loss of scaling law and model collapse, further indicating that model collapse is even more pronounced here, highlighting the challenges in training next generation LLMs. More details and results in Appendix H.
|
Figure 2: **Performance Overview.** The figure depicts the success of various models on the MGSM-JA task, with each of the 250 test problems represented along the x-axis by problem ID. Correct answers are indicated by colored markers at the corresponding positions.
|
**Theorem 1**.: _Under mild assumptions, all reward classes consistent with the Plackett-Luce (and Bradley-Terry in particular) models can be represented with the reparameterization \(r(x,y)=\beta\log\frac{\pi(y|x)}{\pi_{\text{ref}}(y|x)}\) for some model \(\pi(y\mid x)\) and a given reference model \(\pi_{\text{ref}}(y\mid x)\)._
|
**Major Issues**. Although LLMs have achieved key progress in capturing and utilizing knowledge information, they suffer from two major issues as discussed below.
|
Here, we describe the supervised data collection and training of two models, the critic \(\mathcal{C}\) (Section 3.2.1) and the generator \(\mathcal{M}\) (Section 3.2.2).
|
We run the continual pre-training job on one P4d.24xlarge instance through AWS SageMaker. As the model size is moderate, we only use data parallelism via DeepSpeed ZeRO Stage 2 [20] with activation checkpointing enabled. It takes 18 days for FinPythia-6.9B to pre-train and 3 days for FinPythia-1B to pre-train on 24 billion tokens.
|
**Other Quantization Methods.** In the above, we mainly focus on PTQ methods, and next introduce two recent studies that explore efficient fine-tuning methods or QAT methods for quantizing LLMs.
|
February 2024. We have chosen to anonymize the results because the purpose of releasing the v0.5 benchmark is not to draw attention to the strengths or weaknesses of any single model, and the v0.5 benchmark has numerous limitations.
|
+
Footnote †: This bandana makes the perfect addition to every für babies birthday collection! With its vibrant watercolor popsle pattern, your pup will be ready for every summer cookout! _With snaps for security, this bandana is made with love, down to the very last stitch!
|
13. Iris Winnow accepts a position as a war correspondent with the Inridridation Tribune newspaper.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.