Benchmarks have emerged as the central approach for evaluating Large Language Models (LLMs). The research community often relies on a model’s average performance across the test prompts of a benchmark to evaluate the model’s performance. This is consistent with the assumption that the test prompts within a benchmark represent a random sample from some real-world distribution of interest. We note that this is generally not the case; instead, we hold that the distribution of interest varies according to the specific use case. Hence, we analyze the robustness of LLM benchmarks to their underlying distributional assumptions. We find that (1) the correlation in model performance across test prompts is non-random, (2) accounting for correlations across test prompts can change model rankings on major benchmarks, (3) explanatory factors for these correlations include semantic similarity and common LLM failure points.
论文检索
输入标题、作者或关键词，从 9,460 篇学术成果中精准定位
Legal tasks and datasets are often used as benchmarks for the capabilities of language models. However, openly available annotated datasets are rare. In this paper, we introduce AGB-DE, a corpus of 3,764 clauses from German consumer contracts that have been annotated and legally assessed by legal experts. Together with the data, we present a first baseline for the task of detecting potentially void clauses, comparing the performance of an SVM baseline with three fine-tuned open language models and the performance of GPT-3.5. Our results show the challenging nature of the task, with no approach exceeding an F1-score of 0.54. While the fine-tuned models often performed better with regard to precision, GPT-3.5 outperformed the other approaches with regard to recall. An analysis of the errors indicates that one of the main challenges could be the correct interpretation of complex clauses, rather than the decision boundaries of what is permissible and what is not.
Large language models (LLMs) have achieved remarkable performance in natural language understanding and generation tasks. However, they often suffer from limitations such as difficulty in incorporating new knowledge, generating hallucinations, and explaining their reasoning process. To address these challenges, we propose a novel prompting pipeline, named MindMap, that leverages knowledge graphs (KGs) to enhance LLMs’ inference and transparency. Our method enables LLMs to comprehend KG inputs and infer with a combination of implicit and external knowledge. Moreover, our method elicits the mind map of LLMs, which reveals their reasoning pathways based on the ontology of knowledge. We evaluate our method on diverse question & answering tasks, especially in medical domains, and show significant improvements over baselines. We also introduce a new hallucination evaluation benchmark and analyze the effects of different components of our method. Our results demonstrate the effectiveness and robustness of our method in merging knowledge from LLMs and KGs for combined inference.
Automated teaching assistants and chatbots have significant potential to reduce the workload of human instructors, especially for logistics-related question answering, which is important to students yet repetitive for instructors. However, due to privacy concerns, there is a lack of publicly available datasets. We introduce SyllabusQA, an open-source dataset with 63 real course syllabi covering 36 majors, containing 5,078 open-ended course logistics-related question-answer pairs that are diverse in both question types and answer formats. Since many logistics-related questions contain critical information like the date of an exam, it is important to evaluate the factuality of answers. We benchmark several strong baselines on this task, from large language model prompting to retrieval-augmented generation. We introduce Fact-QA, an LLM-based (GPT-4) evaluation metric to evaluate the factuality of predicted answers. We find that despite performing close to humans on traditional metrics of textual similarity, there remains a significant gap between automated approaches and humans in terms of fact precision.
The rapid propagation of misinformation poses substantial risks to public interest. To combat misinformation, large language models (LLMs) are adapted to automatically verify claim credibility. Nevertheless, existing methods heavily rely on the embedded knowledge within LLMs and / or black-box APIs for evidence collection, leading to subpar performance with smaller LLMs or upon unreliable context. In this paper, we propose retrieval augmented fact verification through the synthesis of contrasting arguments (RAFTS). Upon input claims, RAFTS starts with evidence retrieval, where we design a retrieval pipeline to collect and re-rank relevant documents from verifiable sources. Then, RAFTS forms contrastive arguments (i.e., supporting or refuting) conditioned on the retrieved evidence. In addition, RAFTS leverages an embedding model to identify informative demonstrations, followed by in-context prompting to generate the prediction and explanation. Our method effectively retrieves relevant documents as evidence and evaluates arguments from varying perspectives, incorporating nuanced information for fine-grained decision-making. Combined with informative in-context examples as prior, RAFTS achieves significant improvements to supervised and LLM baselines without complex prompts. We demonstrate the effectiveness of our method through extensive experiments, where RAFTS can outperform GPT-based methods with a significantly smaller 7B LLM.
Multiple-choice question answering (MCQA) is often used to evaluate large language models (LLMs). To see if MCQA assesses LLMs as intended, we probe if LLMs can perform MCQA with choices-only prompts, where models must select the correct answer only from the choices. In three MCQA datasets and four LLMs, this prompt bests a majority baseline in 11/12 cases, with up to 0.33 accuracy gain. To help explain this behavior, we conduct an in-depth, black-box analysis on memorization, choice dynamics, and question inference. Our key findings are threefold. First, we find no evidence that the choices-only accuracy stems from memorization alone. Second, priors over individual choices do not fully explain choices-only accuracy, hinting that LLMs use the group dynamics of choices. Third, LLMs have some ability to infer a relevant question from choices, and surprisingly can sometimes even match the original question. We hope to motivate the use of stronger baselines in MCQA benchmarks, the design of robust MCQA datasets, and further efforts to explain LLM decision-making.
Large language models (LLMs) have shown remarkable promise in simulating human language and behavior. This study investigates how integrating persona variables—demographic, social, and behavioral factors—impacts LLMs’ ability to simulate diverse perspectives. We find that persona variables account for <10% variance in annotations in existing subjective NLP datasets. Nonetheless, incorporating persona variables via prompting in LLMs provides modest but statistically significant improvements. Persona prompting is most effective in samples where many annotators disagree, but their disagreements are relatively minor. Notably, we find a linear relationship in our setting: the stronger the correlation between persona variables and human annotations, the more accurate the LLM predictions are using persona prompting. In a zero-shot setting, a powerful 70b model with persona prompting captures 81% of the annotation variance achievable by linear regression trained on ground truth annotations. However, for most subjective NLP datasets, where persona variables have limited explanatory power, the benefits of persona prompting are limited.
Real world deployments of word alignment are almost certain to cover both high and low resource languages. However, the state-of-the-art for this task recommends a different model class depending on the availability of gold alignment training data for a particular language pair. We propose BinaryAlign, a novel word alignment technique based on binary sequence labeling that outperforms existing approaches in both scenarios, offering a unifying approach to the task. Additionally, we vary the specific choice of multilingual foundation model, perform stratified error analysis over alignment error type, and explore the performance of BinaryAlign on non-English language pairs. We make our source code publicly available.
To better analyze informal arguments on public forums, we propose the task of argument explication, which makes explicit a text’s argumentative structure and implicit reasoning by outputting triples of propositions ⟨claim, reason warrant⟩. The three slots, or argument components, are derived from the widely known Toulmin (1958) model of argumentation. While prior research applies Toulmin or related theories to annotate datasets and train supervised models, we develop an effective method to prompt generative large language models (LMs) to output explicitly named argument components proposed by Toulmin by prompting with the theory name (e.g., ‘According to Toulmin model’). We evaluate the outputs’ coverage and validity through a human study and automatic evaluation based on prior argumentation datasets and perform robustness checks over alternative LMs, prompts, and argumentation theories. Finally, we conduct a proof-of-concept case study to extract an interpretable argumentation (hyper)graph from a large corpus of critical public comments on whether to allow the COVID-19 vaccine for children, suggesting future directions for corpus analysis and argument visualization.
In math reasoning with large language models (LLMs), fine-tuning data augmentation by query evolution and diverse reasoning paths is empirically verified effective, profoundly narrowing the gap between open-sourced LLMs and cutting-edge proprietary LLMs. In this paper, we conduct an investigation for such data augmentation in math reasoning and are intended to answer: (1) What strategies of data augmentation are more effective; (2) What is the scaling relationship between the amount of augmented data and model performance; and (3) Can data augmentation incentivize generalization to out-of-domain mathematical reasoning tasks?To this end, we create two new dataset AugGSM8K and AugMATH, by complicating and diversifying the queries and sampling multiple reasoning paths from GSM8K and MATH.We obtained a series of LLMs called MuggleMath by fine-tuning LLaMA models on AugGSM8K and AugMATH. MuggleMath substantially achieves new state-of-the-art on GSM8K and MATH.A log-linear relationship and a segmented log-linear are presented between MuggleMath’s performance and the amount of augmented data on GSM8K and MATH, respectively.We also find that it is weak in out-of-domain math reasoning generalization from AugGSM8K to MATH and from AugMATH to GSM8K, which suggests that augmenting queries that cover a broader range of subjects is more beneficial for generalization.
We study whether Large Language Models (LLMs) latently perform multi-hop reasoning with complex prompts such as “The mother of the singer of ‘Superstition’ is”. We look for evidence of a latent reasoning pathway where an LLM (1) latently identifies “the singer of ‘Superstition’” as Stevie Wonder, the bridge entity, and (2) uses its knowledge of Stevie Wonder’s mother to complete the prompt. We analyze these two hops individually and consider their co-occurrence as indicative of latent multi-hop reasoning. For the first hop, we test if changing the prompt to indirectly mention the bridge entity instead of any other entity increases the LLM’s internal recall of the bridge entity. For the second hop, we test if increasing this recall causes the LLM to better utilize what it knows about the bridge entity. We find strong evidence of latent multi-hop reasoning for the prompts of certain relation types, with the reasoning pathway used in more than 80% of the prompts. However, the utilization is highly contextual, varying across different types of prompts. Also, on average, the evidence for the second hop and the full multi-hop traversal is rather moderate and only substantial for the first hop. Moreover, we find a clear scaling trend with increasing model size for the first hop of reasoning but not for the second hop. Our experimental findings suggest potential challenges and opportunities for future development and applications of LLMs.
OWSM-CTC: An Open Encoder-Only Speech Foundation Model for Speech Recognition, Translation, and Language Identification
PDF ↗There has been an increasing interest in large speech models that can perform multiple tasks in a single model. Such models usually adopt an encoder-decoder or decoder-only architecture due to their popularity and good performance in many domains. However, autoregressive models can be slower during inference compared to non-autoregressive models and also have potential risks of hallucination. Though prior studies observed promising results of non-autoregressive models for certain tasks at small scales, it remains unclear if they can be scaled to speech-to-text generation in diverse languages and tasks. Inspired by the Open Whisper-style Speech Model (OWSM) project, we propose OWSM-CTC, a novel encoder-only speech foundation model based on Connectionist Temporal Classification (CTC). It is trained on 180k hours of public audio data for multilingual automatic speech recognition (ASR), speech translation (ST), and language identification (LID). Compared to encoder-decoder OWSM, our OWSM-CTC achieves competitive results on ASR and up to 24% relative improvement on ST, while it is more robust and 3 to 4 times faster for inference. OWSM-CTC also improves the long-form ASR result with 20x speed-up.We will publicly release our code, pre-trained model, and training logs to promote open science in speech foundation models.
Structured Sentiment Analysis (SSA) was cast as a problem of bi-lexical dependency graph parsing by prior studies.Multiple formulations have been proposed to construct the graph, which share several intrinsic drawbacks:(1) The internal structures of spans are neglected, thus only the boundary tokens of spans are used for relation prediction and span recognition, thus hindering the model’s expressiveness;(2) Long spans occupy a significant proportion in the SSA datasets, which further exacerbates the problem of internal structure neglect.In this paper, we treat the SSA task as a dependency parsing task on partially-observed dependency trees, regarding flat spans without determined tree annotations as latent subtrees to consider internal structures of spans.We propose a two-stage parsing method and leverage TreeCRFs with a novel constrained inside algorithm to model latent structures explicitly, which also takes advantages of joint scoring graph arcs and headed spans for global optimization and inference. Results of extensive experiments on five benchmark datasets reveal that our method performs significantly better than all previous bi-lexical methods, achieving new state-of-the-art.
Modern natural language generation systems with Large Language Models (LLMs) exhibit the capability to generate a plausible summary of multiple documents; however, it is uncertain if they truly possess the capability of information consolidation to generate summaries, especially on documents with opinionated information. We focus on meta-review generation, a form of sentiment summarisation for the scientific domain. To make scientific sentiment summarization more grounded, we hypothesize that human meta-reviewers follow a three-layer framework of sentiment consolidation to write meta-reviews. Based on the framework, we propose novel prompting methods for LLMs to generate meta-reviews and evaluation metrics to assess the quality of generated meta-reviews. Our framework is validated empirically as we find that prompting LLMs based on the framework — compared with prompting them with simple instructions — generates better meta-reviews.
We introduce a new unsupervised text embedding method, Meta-Task Prompting with Explicit One-Word Limitation (MetaEOL), for generating high-quality sentence embeddings from Large Language Models (LLMs) without the need for model fine-tuning. Leveraging meta-task prompting, MetaEOL guides LLMs to produce embeddings through a series of carefully designed prompts that address multiple representational aspects. Our comprehensive experiments demonstrate that embeddings averaged from various meta-tasks are versatile embeddings that yield competitive performance on Semantic Textual Similarity (STS) benchmarks and excel in downstream tasks, surpassing contrastive-trained models. Our findings suggest a new scaling law, offering a versatile and resource-efficient approach for embedding generation across diverse scenarios.
SymKGQA: Few-Shot Knowledge Graph Question Answering via Symbolic Program Generation and Execution
PDF ↗Semantic Parsing of natural language questions into their executable logical form (LF) has shown state-of-the-art (SOTA) performance for Knowledge Graph Question Answering (KGQA). However, these methods are not applicable for real-world applications, due to lack of KG-specific training data. Recent advances in the capabilities of Large Language Models (LLMs) has led towards generating low-level LFs such as SPARQL and S-Expression in a few-shot setting. Unfortunately, these methods: (1) are limited to the knowledge of underlying LLM about the LF, (2) performs inferior for the harder complex benchmarks such as KQA Pro, (3) suffers while grounding the generated LF to a specific Knowledge Graph. Recently, a new LF called KoPL has been introduced that explicitly models complex reasoning process step-by-step in a symbolic manner and has shown SOTA on KQA Pro in fully-supervised setting. Inspired by this, we propose SymKGQA framework that generates step-by-step Symbolic LF i.e., KoPL in a few-shot in-context learning setting using LLM. Our framework is not dependent on pre-trained information of LLM about KoPL. We further build a Retrieval-Augmented Generation based Question-Aware Contextual KoPL (QUACK) resolver to ground the generated LF. Our experiments with different LLMs and few-shot settings demonstrate that SymKGQA outperforms all other few-shot and even many of the fully-supervised KGQA approaches.
Quantization and pruning form the foundation of compression for neural networks, enabling efficient inference for large language models (LLMs). Recently, various quantization and pruning techniques have demonstrated remarkable performance in a post-training setting. They rely upon calibration data, a small set of unlabeled examples that are used to generate layer activations. However, no prior work has systematically investigated how the calibration data impacts the effectiveness of model compression methods. In this paper, we present the first extensive empirical study on the effect of calibration data upon LLM performance. We trial a variety of quantization and pruning methods, datasets, tasks, and models. Surprisingly, we find substantial variations in downstream task performance, contrasting existing work that suggests a greater level of robustness to the calibration data. Finally, we make a series of recommendations for the effective use of calibration data in LLM quantization and pruning.
In the field of speech synthesis, there is a growing emphasis on employing multimodal speech to enhance robustness. A key challenge in this area is the scarcity of datasets that pair audio with corresponding video. We employ a methodology that incorporates modality alignment during the pre-training phase on multimodal datasets, uniquely facilitating zero-shot generalization through the process of freezing the video modality feature extraction component and the encoder module within the pretrained weights, thereby enabling effective cross-modal and cross-lingual transfer. We have named this method ‘Uni-Dubbing’. Our method finely tunes with both multimodal and single-modality audio data. In multimodal scenarios, it achieves a reduced word error rate (WER) of 31.73%, surpassing the previous best of 33.9%. It also excels in metrics like tone quality and synchronization. With single-modality audio, it achieves a WER of 36.08%, demonstrating adaptability to limited data. Its domain generalization capabilities are proven across various language tasks in video translation and audio generation. Trained on 433 hours of audio data, it surpasses techniques using 200 hours of audiovisual data. The code and demo are available at https://diracer.github.io/unidubbing.
Recent studies have shown that integrating constructional information can improve the performance of pre-trained language models (PLMs) in natural language understanding. However, exploration into leveraging constructional information to enhance generative language models for natural language generation has been limited. Additionally, probing studies indicate that PLMs primarily grasp the syntactic structure of constructions but struggle to capture their semantics. In this work, we encode constructions as inductive biases to explicitly embed constructional semantics and guide the generation process. We begin by presenting a construction grammar induction framework designed to automatically identify constructions from corpora. Subsequently, we propose the Construction-Enhanced Language Model (CoELM). It introduces a construction-guided language modeling approach that employs a dynamic sequence reassembly strategy during pre-training. Extensive experiments have demonstrated the superiority of CoELM across various benchmarks.
Comparative reasoning plays a crucial role in predicting text preferences; however, large language models (LLMs) often demonstrate inconsistencies in their reasoning, leading to incorrect preference predictions. While approaches like Chain-of-Thought improve accuracy in many settings, they struggle to consistently distinguish the similarities and differences of complex texts. We introduce SC^2, a model that prompts LLMs to predict text preferences by generating structured intermediate comparisons. SC^2 begins by proposing aspects for comparison, followed by generating textual comparisons under each aspect. We select consistent comparisons with a pairwise comparator that ensures each comparison of a given aspect clearly distinguishes differences between texts, significantly reducing hallucination and improving consistency. Our empirical studies across various NLP tasks, including summarization, retrieval, and automatic rating, demonstrate that SC^2‘s enhanced performance in text preference prediction is significant.