We design probes trained on the internal representations of a transformer language model to predict its hallucinatory behavior on three grounded generation tasks. To train the probes, we annotate for span-level hallucination on both sampled (organic) and manually edited (synthetic) reference outputs. Our probes are narrowly trained and we find that they are sensitive to their training domain: they generalize poorly from one task to another or from synthetic to organic hallucinations. However, on in-domain data, they can reliably detect hallucinations at many transformer layers, achieving 95% of their peak performance as early as layer 4. Here, probing proves accurate for evaluating hallucination, outperforming several contemporary baselines and even surpassing an expert human annotator in response-level detection F1. Similarly, on span-level labeling, probes are on par or better than the expert annotator on two out of three generation tasks. Overall, we find that probing is a feasible and efficient alternative to language model hallucination evaluation when model states are available.
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
Planning, Creation, Usage: Benchmarking LLMs for Comprehensive Tool Utilization in Real-World Complex Scenarios
PDF ↗The recent trend of using Large Language Models (LLMs) as tool agents in real-world applications underscores the necessity for comprehensive evaluations of their capabilities, particularly in complex scenarios involving planning, creating, and using tools. However, existing benchmarks typically focus on simple synthesized queries that do not reflect real-world complexity, thereby offering limited perspectives in evaluating tool utilization. To address this issue, we present UltraTool, a novel benchmark designed to improve and evaluate LLMs’ ability in tool utilization within real-world scenarios. UltraTool focuses on the entire process of using tools - from planning and creating to applying them in complex tasks. It emphasizes real-world complexities, demanding accurate, multi-step planning for effective problem-solving. A key feature of UltraTool is its independent evaluation of planning with natural language, which happens before tool usage and simplifies the task solving by mapping out the intermediate steps. Thus, unlike previous work, it eliminates the restriction of pre-defined toolset. Through extensive experiments on various LLMs, we offer novel insights into the evaluation of capabilities of LLMs in tool utilization, thereby contributing a fresh perspective to this rapidly evolving field. The benchmark is publicly available at https://github.com/JoeYing1019/UltraTool.
Controlling the attribute intensity of text generation is crucial across scenarios (e.g., writing conciseness, chatting emotion, and explanation clarity). The remarkable capabilities of large language models (LLMs) have revolutionized text generation, prompting us to explore such smooth control of LLM generation. Specifically, we propose metrics to assess the range, calibration, and consistency of the generated text’s attribute intensity in response to varying control values, as well as its relevance to the intended context. To quantify the attribute intensity and context relevance, we leverage an Elo rating system and GPT4, respectively, both renowned for their robust alignment with human judgment. We look into two viable training-free methods for achieving smooth control of LLMs: (1) Prompting with semantic shifters, and (2) Modifying internal model representations. The evaluations of these two methods are conducted on 5 different attributes with various models.
Large language models (LLMs), despite their breakthroughs on many challenging benchmark tasks, prefer to generate verbose responses and lack the controllability of output complexity, which is usually preferred by human users in practice. In this paper, we study how to precisely control multiple linguistic complexities of LLM output by finetuning using off-the-shelf data. To this end, we propose multi-control tuning (MCTune), which includes multiple linguistic complexity values of ground-truth responses as controls in the input for instruction tuning. We finetune LLaMA2-7B on Alpaca-GPT4 and WizardLM datasets. Evaluations on widely used benchmarks demonstrate that our method does not only improve LLMs’ multi-complexity controllability substantially but also retains or even enhances the quality of the responses as a side benefit.
Knowledge graph reasoning is an important problem for knowledge graphs. In this paper, we propose a novel and principled framework called RulE (stands for Rule Embedding) to effectively leverage logical rules to enhance KG reasoning. Unlike knowledge graph embedding methods, RulE learns rule embeddings from existing triplets and first-order rules by jointly representing entities, relations and logical rules in a unified embedding space. Based on the learned rule embeddings, a confidence score can be calculated for each rule, reflecting its consistency with the observed triplets. This allows us to perform logical rule inference in a soft way, thus alleviating the brittleness of logic. On the other hand, RulE injects prior logical rule information into the embedding space, enriching and regularizing the entity/relation embeddings. This makes KGE alone perform better too. RulE is conceptually simple and empirically effective. We conduct extensive experiments to verify each component of RulE.Results on multiple benchmarks reveal that our model outperforms the majority of existing embedding-based and rule-based approaches.
Legal Judgment Reimagined: PredEx and the Rise of Intelligent AI Interpretation in Indian Courts
PDF ↗In the era of Large Language Models (LLMs), predicting judicial outcomes poses significant challenges due to the complexity of legal proceedings and the scarcity of expert-annotated datasets. Addressing this, we introduce Prediction with Explanation (PredEx), the largest expert-annotated dataset for legal judgment prediction and explanation in the Indian context, featuring over 15,000 annotations. This groundbreaking corpus significantly enhances the training and evaluation of AI models in legal analysis, with innovations including the application of instruction tuning to LLMs. This method has markedly improved the predictive accuracy and explanatory depth of these models for legal judgments. We employed various transformer-based models, tailored for both general and Indian legal contexts. Through rigorous lexical, semantic, and expert assessments, our models effectively leverage PredEx to provide precise predictions and meaningful explanations, establishing it as a valuable benchmark for both the legal profession and the NLP community.
Call Me When Necessary: LLMs can Efficiently and Faithfully Reason over Structured Environments
PDF ↗Large Language Models (LLMs) have shown potential in reasoning over structured environments, e.g., knowledge graphs and tables. Such tasks typically require multi-hop reasoning, i.e., match natural language utterance with instances in the environment. Previous works adopt LLMs to incrementally build a reasoning path, where LLMs either invoke tools or pick up items by step-by-step interacting with the environment. We propose Reasoning-Path-Editing (Readi), a novel framework where LLMs can efficiently and faithfully reason over structured environments. In Readi, LLMs initially generate a reasoning path given a query, and edit the path only when necessary. We instantiate the path on structured environments and provide feedback to edit the path if anything goes wrong. Experimental results on three KGQA and two TableQA datasets show the effectiveness of Readi, significantly surpassing previous LLM-based methods (by 9.1% Hit@1 on WebQSP, 12.4% on MQA-3H and 9.5% on WTQ), comparable with state-of-the-art fine-tuned methods (67% on CWQ and 74.7% on WebQSP) and substantially boosting the vanilla LLMs (by 14.9% on CWQ). Our code will be available on https://aka.ms/readi.
Are Female Carpenters like Blue Bananas? A Corpus Investigation of Occupation Gender Typicality
PDF ↗People tend to use language to mention surprising properties of events: for example, when a banana is blue, we are more likely to mention color than when it is yellow. This fact is taken to suggest that yellowness is somehow a typical feature of bananas, and blueness is exceptional. Similar to how a yellow color is typical of bananas, there may also be genders that are typical of occupations. In this work, we explore this question using information theoretic techniques coupled with corpus statistic analysis. In two distinct large corpora, we do not find strong evidence that occupations and gender display the same patterns of mentioning as do bananas and color. Instead, we find that gender mentioning is correlated with femaleness of occupation in particular, suggesting perhaps that woman-dominated occupations are seen as somehow “more gendered” than male-dominated ones, and thereby they encourage more gender mentioning overall.
Training Large Language Models (LLMs) to process extensive context lengths incurs prohibitive computational costs. Prevailing techniques for extending context capabilities in LLMs typically require not only additional training procedures but also access to datasets with long context (e.g., sequences of 32K tokens), presupposing substantial GPU expenditures. To address the aforementioned issues, we introduce a novel solution named Efficient and Extreme length extension for Large Language Models (E2-LLM). E2-LLM entails a singular training process over considerably short sequences (e.g., 4K tokens), which greatly mitigates the cost of continual-pretraining or fine-tuning. Within the training phase, we incorporate a dual augmentation strategy with Rotary Position Embeddings (RoPE) that adjusts the scale and position indices across distinct training samples. E 2 -LLM is meticulously designed to enhance the model’s robustness to diverse relative positions. The experimental results on multiple benchmark datasets demonstrate the superior performance of E 2 -LLM on demanding tasks of processing long contexts.
We release a multi-accent dataset and propose speech-programming and gradient reversal classifier to improve the generalization.Abstract: Speech-to-SQL (S2SQL) aims to convert spoken questions into SQL queries given relational databases, which has been traditionally implemented in a cascaded manner while facing the following challenges: 1) model training is faced with the major issue of data scarcity, where limited parallel data is available; and 2) the systems should be robust enough to handle diverse out-of-domain speech samples that differ from the source data. In this work, we propose the direct generalizable speech-to-SQL parsing model Wav2SQL which avoids error compounding across cascaded systems. Specifically, 1) to accelerate speech-driven SQL parsing research in the community, we release a large-scale and multi-accent dataset MASpider; 2) leveraging the recent progress in the large-scale pre-training, we show that it alleviates the data scarcity issue and allow for direct speech-to-SQL parsing; and 3) we include the speech re-programming and gradient reversal classifier techniques to reduce acoustic variance and learned style-agnostic representation, improving generalization to unseen out-of-domain custom data. Experimental results demonstrate that Wav2SQL avoids error compounding and achieves state-of-the-art results by up to 4.7% accuracy improvement over the baseline.
Recently, large multi-modal models (LMMs) have emerged with the capacity to perform vision tasks such as captioning and visual question answering (VQA) with unprecedented accuracy. Applications such as helping the blind or visually impaired have a critical need for precise answers. It is specially important for models to be well calibrated and be able to quantify their uncertainty in order to selectively decide when to answer and when to abstain or ask for clarifications. We perform the first in-depth analysis of calibration methods and metrics for VQA with in-context learning LMMs. Studying VQA on two answerability benchmarks, we show that the likelihood score of visually grounded models is better calibrated than in their text-only counterparts for in-context learning, where sampling based methods are generally superior, but no clear winner arises. We propose Avg BLEU, a calibration score combining the benefits of both sampling and likelihood methods across modalities.
The objective of the Causal Emotion Entailment (CEE) task is to identify the causes of the target emotional utterances in a given conversation. Most existing studies have focused on a fine-tuning paradigm based on a pretrained model, e.g., the BERT model. However, there are gaps between the pretrained task and the CEE task. Although a pretrained model enhances contextual comprehension to some extent, it cannot acquire specific knowledge that is relevant to the CEE task. In addition, in a typical CEE task, there are peculiarities in the distribution of the positions with different emotion types of emotion utterances and cause utterances in conversations. Existing methods employ a fixed-size window to capture the relationship between neighboring conversations; however, these methods ignore the specific semantic associations between emotions and cause utterances. To address these issues, we propose the Position-oriented Prompt-tuning (POP-CEE) model to solve the CEE task in an end-to-end manner. Specifically, we can model the CEE task by designing prompts with multiple unified goals and by exploring the positional relationship between emotion and cause utterances using a position constraint module. Experimental results demonstrate that the proposed POP-CEE model achieves state-of-the-art performance on a benchmark dataset. Ourcode and data can be found at: https://github.com/Zh0uzh/POP-CEE.
Large Language Models (LLMs) have demonstrated exceptional coding capability. However, as another critical component of programming proficiency, the debugging capability of LLMs remains relatively unexplored. Previous evaluations of LLMs’ debugging ability are significantly limited by the risk of data leakage, the scale of the dataset, and the variety of tested bugs. To overcome these deficiencies, we introduce ‘DebugBench’, an LLM debugging benchmark consisting of 4,253 instances. It covers four major bug categories and 18 minor types in C++, Java, and Python. To construct DebugBench, we collect code snippets from the LeetCode community, implant bugs into source data with GPT-4, and assure rigorous quality checks. We evaluate two commercial and four open-source models in a zero-shot scenario. We find that (1) while closed-source models exhibit inferior debugging performance compared to humans, open-source models relatively lower pass rate scores; (2) the complexity of debugging notably fluctuates depending on the bug category; (3) incorporating runtime feedback has a clear impact on debugging performance which is not always helpful. As an extension, we also compare LLM debugging and code generation, revealing a strong correlation between them for closed-source models. These findings will benefit the development of LLMs in debugging.
Your Vision-Language Model Itself Is a Strong Filter: Towards High-Quality Instruction Tuning with Data Selection
PDF ↗Data selection in instruction tuning emerges as a pivotal process for acquiring high-quality data and training instruction-following large language models (LLMs), but it is still a new and unexplored research area for vision-language models (VLMs). Existing data selection approaches on LLMs either rely on single unreliable scores, or use downstream tasks for selection, which is time-consuming and can lead to potential over-fitting on the chosen evaluation datasets. To address this challenge, we introduce a novel dataset selection method, Self-Filter, that utilizes the VLM itself as a filter. This approach is inspired by the observation that VLMs benefit from training with the most challenging instructions. Self-Filter operates in two stages. In the first stage, we devise a scoring network to evaluate the difficulty of training instructions, which is co-trained with the VLM. In the second stage, we use the trained score net to measure the difficulty of each instruction, select the most challenging samples, and penalize similar samples to encourage diversity. Comprehensive experiments on LLaVA and MiniGPT-4 show that Self-Filter can reach better results compared to full data settings with merely about 15% samples, and can achieve superior performance against competitive baselines.
In this survey, we provide a systematic review of recent work on modelling morality in text, an area of research that has garnered increasing attention in recent years. Our survey is motivated by the importance of modelling decisions on the created resources, the models trained on these resources and the analyses that result from the models’ predictions. We review work at the interface of NLP, Computational Social Science and Psychology and give an overview of the different goals and research questions addressed in the papers, their underlying theoretical backgrounds and the methods that have been applied to pursue these goals. We then identify and discuss challenges and research gaps, such as the lack of a theoretical framework underlying the operationalisation of morality in text, the low IAA reported for manyhuman-annotated resulting resources and the lack of validation of newly proposed resources and analyses.
Recent work by Hewitt et al. (2020) provides an interpretation of the empirical success of recurrent neural networks (RNNs) as language models (LMs). It shows that RNNs can efficiently represent bounded hierarchical structures that are prevalent in human language.This suggests that RNNs’ success might be linked to their ability to model hierarchy. However, a closer inspection of hewitt-etal-2020-rnns construction shows that it is not inherently limited to hierarchical structures. This poses a natural question: What other classes of LMs RNNs can efficiently represent? To this end, we generalize Hewitt et al.’s (2020) construction and show that RNNs can efficiently represent a larger class of LMs than previously claimed—specifically, those that can be represented by a pushdown automaton with a bounded stack and a specific stack update function. Altogether, the efficiency of representing this diverse class of LMs with RNN LMs suggests novel interpretations of their inductive bias.
Multimodal entity linking (MEL) aims to link ambiguous mentions in multimodal contexts to entities in a multimodal knowledge graph. A pivotal challenge is to fully leverage multi-element correlations between mentions and entities to bridge modality gap and enable fine-grained semantic matching. Existing methods attempt several local correlative mechanisms, relying heavily on the automatically learned attention weights, which may over-concentrate on partial correlations. To mitigate this issue, we formulate the correlation assignment problem as an optimal transport (OT) problem, and propose a novel MEL framework, namely OT-MEL, with OT-guided correlation assignment. Thereby, we exploit the correlation between multimodal features to enhance multimodal fusion, and the correlation between mentions and entities to enhance fine-grained matching. To accelerate model prediction, we further leverage knowledge distillation to transfer OT assignment knowledge to attention mechanism. Experimental results show that our model significantly outperforms previous state-of-the-art baselines and confirm the effectiveness of the OT-guided correlation assignment.
Transformers demonstrate impressive performance on a range of reasoning benchmarks. To evaluate the degree to which these abilities are a result of actual reasoning, existing work has focused on developing sophisticated benchmarks for behavioral studies. However, these studies do not provide insights into the internal mechanisms driving the observed capabilities. To improve our understanding of the internal mechanisms of transformers, we present a comprehensive mechanistic analysis of a transformer trained on a synthetic reasoning task. We identify a set of interpretable mechanisms the model uses to solve the task, and validate our findings using correlational and causal evidence. Our results suggest that it implements a depth-bounded recurrent mechanisms that operates in parallel and stores intermediate results in selected token positions. We anticipate that the motifs we identified in our synthetic setting can provide valuable insights into the broader operating principles of transformers and thus provide a basis for understanding more complex models.
English NLP systems have empirically worse performance for dialects other than Standard American English (SAmE). However, how these discrepancies impact use of language technology by speakers of non-SAmE global Englishes is not well understood. We focus on reducing this gap for South Asian Englishes (SAsE), a macro-group of regional varieties with cumulatively more speakers than SAmE, by surveying SAsE speakers about their interactions with language technology and compare their responses to a control survey of SAmE speakers. SAsE speakers are more likely to recall failures with language technology and more likely to reference specific issues with written language technology than their SAmE counterparts. Furthermore, SAsE speakers indicate that they modify both their lexicon and syntax to make technology work better, but that lexical issues are perceived as the most salient challenge. We then assess whether these issues are pervasive in more recently developed Large Language Models (LLMs), introducing two benchmarks for broader SAsE Lexical and Indian English Syntactic understanding and evaluating 11 families of LLMs on them.