ACL 2024 论文列表

Virtual environments play a key role in benchmarking advances in complex planning and decision-making tasks but are expensive and complicated to build by hand. Can current language models themselves serve as world simulators, correctly predicting how actions change different world states, thus bypassing the need for extensive manual coding? Our goal is to answer this question in the context of text-based simulators. Our approach is to build and use a new benchmark, called ByteSized32-State-Prediction, containing a dataset of text game state transitions and accompanying game tasks. We use this to directly quantify, for the first time, how well LLMs can serve as text-based world simulators. We test GPT-4 on this dataset and find that, despite its impressive performance, it is still an unreliable world simulator without further innovations. This work thus contributes both new insights into current LLM’s capabilities and weaknesses, as well as a novel benchmark to track future progress as new models appear.

Making LLMs speak for different, especially minority groups of people, and generate statements supporting their diverse or even controversial perspectives is critical to creating an inclusive environment. However, existing LLMs lack sufficient controllability to the stance of their generated content, which often contains inconsistent, neutral, or biased statements. In this paper, we improve the controllability of LLMs in generating statements supporting an argument the user defined in the prompt. We find that multi-round debates between two LLMs with opposite stances generate higher-quality and more salient statements for each, which are important training data to improve the controllability of LLMs. Motivated by this, we develop a novel debate & tuning (“DEBATUNE”) pipeline finetuning LLMs to generate the statements obtained via debate. To examine DEBATUNE, we curate the largest dataset of debate topics so far, which covers 710 controversial topics and corresponding arguments for each topic. Evaluations by the GPT-4 judge with a novel controversy controllability metric show that LLMs’ capability of generating diverse perspectives is significantly improved by DEBATUNE. Moreover, such controllability can be generalized to unseen topics, generating high-quality statements supporting controversial arguments.

Large language models (LLMs) have achieved impressive human-like performance across various reasoning tasks. However, their mastery of underlying inferential rules still falls short of human capabilities. To investigate this, we propose a logic scaffolding inferential rule generation framework, to construct an inferential rule base, ULogic, comprising both primitive and compositional rules across five domains. Our analysis of GPT-series models over a rule subset reveals significant gaps in LLMs’ logic understanding compared to human performance, especially in compositional and structural complex rules with certain bias patterns. We further distill these rules into a smaller-scale inference engine for flexible rule generation and enhancing downstream reasoning. Through a multi-judger evaluation, our inference engine proves effective in generating accurate, complex and abstract conclusions and premises, and improve various commonsense reasoning tasks. Overall, our work sheds light on LLMs’ limitations in grasping inferential rule and suggests ways to enhance their logical reasoning abilities .

Large language models (LLMs) have demonstrated striking reasoning capability. Recent works have shown the benefits to LLMs from fine-tuning golden-standard Chain-of-Thought (CoT) rationales or using them as correct examples in few-shot prompting. While humans can indeed imitate correct examples, learning from our mistakes is another vital aspect of human cognition. Hence, a question naturally arises: can LLMs learn and benefit from their mistakes, especially for their reasoning?This study investigates this problem from both the prompting and model-tuning perspectives. We begin by introducing CoTErrorSet, a new benchmark with 609,432 questions, each designed with both correct and error references, and demonstrating the types and reasons for making such mistakes. To explore the effectiveness of those mistakes, we design two methods: (1) Self-rethinking prompting guides LLMs to rethink whether they have made similar previous mistakes; and (2) Mistake tuning involves finetuning models in both correct and incorrect reasoning domains, rather than only tuning models to learn ground truth in traditional methodology. We conduct a series of experiments to prove LLMs can obtain benefits from mistakes in both directions. Our two methods offer potentially cost-effective strategies by leveraging errors to enhance reasoning capabilities, which costs significantly less than creating meticulously hand-crafted golden references. We ultimately make a thorough analysis of the reasons behind LLMs’ errors, which provides directions that future research needs to overcome. CoTErrorSet will be published soon on https://github.com/YookiTong/Learn-from-Mistakes-CotErrorSet.

Metaphors detection, as an important task in the field of NLP, has been receiving sustained academic attention in recent years. Current researches focus supervised metaphors detection systems, which usually require large-scale, high-quality labeled data support. The emerge of large language models (e.g., ChatGPT) has made many NLP tasks (e.g., automatic summarization and dialogue systems) a qualitative leap. However, it is worth noting that the use of ChatGPT for unsupervised metaphors detection is often challenged with less-than-expected performance. Therefore, the aim of our work is to explore how to bootstrap and combine ChatGPT by detecting the most prevalent verb metaphors among metaphors. Our approach first utilizes ChatGPT to obtain literal collocations of target verbs and subject-object pairs of verbs in the text to be detected. Subsequently, these literal collocations and subject-object pairs are mapped to the same set of topics, and finally the verb metaphors are detected through the analysis of entailment relations. The experimental results show that our method achieves the best performance on the unsupervised verb metaphors detection task compared to existing unsupervised methods or direct prediction using ChatGPT. Our code is available at https://github.com/VILAN-Lab/Unsupervised-Metaphor-Detection.

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.

As large language models (LLMs) are increasingly deployed in user-facing applications, building trust and maintaining safety by accurately quantifying a model’s confidence in its prediction becomes even more important. However, finding effective ways to calibrate LLMs—especially when the only interface to the models is their generated text—remains a challenge. We propose APRICOT (Auxiliary prediction of confidence targets): A method to set confidence targets and train an additional model that predicts an LLM’s confidence based on its textual input and output alone. This approach has several advantages: It is conceptually simple, does not require access to the target model beyond its output, does not interfere with the language generation, and has a multitude of potential usages, for instance by verbalizing the predicted confidence or using it to re-prompting the LLM to accurately reflecting its uncertainty. We show how our approach performs competitively in terms of calibration error for white-box and black-box LLMs on closed-book question-answering to detect incorrect LLM answers.

Large-scale deployment of generative AI tools often depends on costly API calls to a Large Language Model (LLM) to fulfil user queries, a process that also exposes the request stream to external providers. To curtail the frequency of these calls, one can employ a local smaller language model -a student- which is continuously trained on the responses of the LLM. This student gradually gains proficiency in independently handling an increasing number of user requests, a process we term neural caching. The crucial element in neural caching is a policy that decides which requests should be processed by the student alone and which should be redirected to the LLM, subsequently aiding the student’s learning. In this study, we focus on classification tasks, and we consider a range of classic Active Learning-based selection criteria as the policy. Our experiments suggest that Margin Sampling and Query by Committee bring consistent benefits over other policies and baselines across tasks and budgets.

In this work, we introduce Context-Aware MultiModal Learner (CaMML), for tuning large multimodal models (LMMs). CaMML, a lightweight module, is crafted to seamlessly integrate multimodal contextual samples into large models, thereby empowering the model to derive knowledge from analogous, domain-specific, up-to-date information and make grounded inferences. Importantly, CaMML is highly scalable and can efficiently handle lengthy multimodal context examples owing to its hierarchical design. Based on CaMML, we have developed two multimodal models, CaMML-7B and CaMML-13B, that have shown exceptional performance across an array of benchmark datasets for multimodal tasks. Remarkably, CaMML-13B achieves the state-of-the-art performance on over ten widely recognized multimodal benchmark datasets, surpassing LLaVA-1.5 (13B) with a noticeable margin, without integration of any external resources. Moreover, we have conducted extensive ablative studies to inspect the inner workings of CaMML and performed qualitative analyses to showcase its effectiveness in handling real-world challenging cases. Code and models are available at: https://github.com/amazon-science/camml.

Grounded generation aims to equip language models (LMs) with the ability to produce more credible and accountable responses by accurately citing verifiable sources. However, existing methods, by either feeding LMs with raw or preprocessed materials, remain prone to errors. To address this, we introduce CaLM, a novel verification framework. CaLM leverages the insight that a robust grounded response should be consistent with information derived solely from its cited sources. Our framework empowers smaller LMs, which rely less on parametric memory and excel at processing relevant information given a query, to validate the output of larger LMs. Larger LM responses that closely align with the smaller LMs’ output, which relies exclusively on cited documents, are verified. Responses showing discrepancies are iteratively refined through a feedback loop. Experiments on three open-domain question-answering datasets demonstrate significant performance gains of 1.5% to 7% absolute average without any required model fine-tuning.

Assessing the capabilities of large language models (LLMs) as agents in decision making and operational tasks is crucial for the development of LLM-as-agent service. We propose CToolEval, a benchmark designed to evaluate LLMs in the context of Chinese societal applications, featuring 398 APIs across 27 widely-used Apps (e.g., Apps for shopping, map, music, travel, etc.) that cover 14 domains. We further present an evaluation framework that simulates real-life scenarios, to facilitate the assessment of tool invocation ability of LLMs for tool learning and task completion ability for user interation. Our extensive experiments with CToolEval evaluate 11 LLMs, revealing that while GPT-3.5-turbo excels in tool invocation, Chinese LLMs usually struggle with issues like hallucination and a lack of comprehensive tool understanding. Our findings highlight the need for further refinement in decision-making capabilities of LLMs, offering insights into bridging the gap between current functionalities and agent-level performance. To promote further research for LLMs to fully act as reliable agents in complex, real-world situations, we release our data and codes at https://github.com/tjunlp-lab/CToolEval.

Direct speech-to-speech translation (S2ST) has achieved impressive translation quality, but it often faces the challenge of slow decoding due to the considerable length of speech sequences. Recently, some research has turned to non-autoregressive (NAR) models to expedite decoding, yet the translation quality typically lags behind autoregressive (AR) models significantly. In this paper, we investigate the performance of CTC-based NAR models in S2ST, as these models have shown impressive results in machine translation. Experimental results demonstrate that by combining pretraining, knowledge distillation, and advanced NAR training techniques such as glancing training and non-monotonic latent alignments, CTC-based NAR models achieve translation quality comparable to the AR model, while preserving up to 26.81× decoding speedup.

In this paper, we present CSCD-NS, the first Chinese spelling check (CSC) dataset designed for native speakers, containing 40,000 samples from a Chinese social platform. Compared with existing CSC datasets aimed at Chinese learners, CSCD-NS is ten times larger in scale and exhibits a distinct error distribution, with a significantly higher proportion of word-level errors. To further enhance the data resource, we propose a novel method that simulates the input process through an input method, generating large-scale and high-quality pseudo data that closely resembles the actual error distribution and outperforms existing methods. Moreover, we investigate the performance of various models in this scenario, including large language models (LLMs), such as ChatGPT. The result indicates that generative models underperform BERT-like classification models due to strict length and pronunciation constraints. The high prevalence of word-level errors also makes CSC for native speakers challenging enough, leaving substantial room for improvement.

It is imperative to ensure the stability of every prediction made by a language model; that is, a language’s prediction should remain consistent despite minor input variations, like word substitutions. In this paper, we investigate the problem of certifying a language model’s robustness against Universal Text Perturbations (UTPs), which have been widely used in universal adversarial attacks and backdoor attacks. Existing certified robustness based on random smoothing has shown considerable promise in certifying the input-specific text perturbations (ISTPs), operating under the assumption that any random alteration of a sample’s clean or adversarial words would negate the impact of sample-wise perturbations. However, with UTPs, masking only the adversarial words can eliminate the attack. A naive method is to simply increase the masking ratio and the likelihood of masking attack tokens, but it leads to a significant reduction in both certified accuracy and the certified radius due to input corruption by extensive masking. To solve this challenge, we introduce a novel approach, the superior prompt search method, designed to identify a superior prompt that maintains higher certified accuracy under extensive masking. Additionally, we theoretically motivate why ensembles are a particularly suitable choice as base prompts for random smoothing. The method is denoted by superior prompt ensembling technique. We also empirically confirm this technique, obtaining state-of-the-art results in multiple settings. These methodologies, for the first time, enable high certified accuracy against both UTPs and ISTPs. The source code of CR-UTP is available at https://github.com/UCF-ML-Research/CR-UTP.

Concept reasoning is an important capability for models to understand the world. However, the existing datasets, such as concept extraction and concept generation, suffer from modeledge leakage and context leakage. To address these limitations, we construct a dataset of concept reasoning for large language models (CR-LLM) with modeledge leakage prevention and context leakage prevention, which consists of 2,167 samples and covers different concept types. In addition, we propose a hybrid reasoning method, consisting of inductive reasoning, deductive reasoning and a controller. This method allows large language models to adaptively select the optimal reasoning method for each input sample. Finally, we conduct extensive experiments on CR-LLM using different models and methods. The results show that existing large language models and reasoning methods perform sub-optimally in the concept reasoning task. In contrast, our proposed method significantly improves the capabilities, achieving a 7% increase in accuracy compared to CoT and demonstrating better granularity. We release CR-LLM and code at https://github.com/Nianqi-Li/Concept-Reasoning-for-LLMs.

The fast-growing large scale language models are delivering unprecedented performance on almost all natural language processing tasks. However, the effectiveness of large language models are reliant on an exponentially increasing number of parameters. The overwhelming computation complexity incurs a high inference latency that negatively affects user experience. Existing methods to improve inference efficiency, such as tensor parallelism and quantization, target to reduce per-layer computing latency, yet overlook the cumulative latency due to the number of layers. Recent works on reducing the cumulative latency through layer removing, however, lead to significant performance drop. Motivated by the similarity of inputs among adjacent layers, we propose to identify quasi-independent layers, which can be concurrently computed to significantly decrease inference latency. We also introduce a bypassing technique to mitigate the effect of information loss. Empirical experiments of the proposed approach on the LLaMA models confirm that Concurrent Computation of Quasi-Independent Layers (CQIL) can reduce latency by up to 48.3% on LLaMA-33B, while maintaining a close level of performance.

Using large language models (LLMs) to assist psychological counseling is a significant but challenging task at present. Attempts have been made on improving empathetic conversations or acting as effective assistants in the treatment with LLMs. However, the existing datasets lack consulting knowledge, resulting in LLMs lacking professional consulting competence. Moreover, how to automatically evaluate multi-turn dialogues within the counseling process remains an understudied area. To bridge the gap, we propose CPsyCoun, a report-based multi-turn dialogue reconstruction and evaluation framework for Chinese psychological counseling. To fully exploit psychological counseling reports, a two-phase approach is devised to construct high-quality dialogues while a comprehensive evaluation benchmark is developed for the effective automatic evaluation of multi-turn psychological consultations. Competitive experimental results demonstrate the effectiveness of our proposed framework in psychological counseling. We open-source the datasets and model for future research.

Assessing the quality of summarizers poses significant challenges—gold summaries are hard to obtain and their suitability depends on the use context of the summarization system. Who is the user of the system, and what do they intend to do with the summary? In response, we propose a novel task-oriented evaluation approach that assesses summarizers based on their capacity to produce summaries while preserving task outcomes. We theoretically establish both a lower and upper bound on the expected error rate of these tasks, which depends on the mutual information between source texts and generated summaries. We introduce COSMIC, a practical implementation of this metric, and demonstrate its strong correlation with human judgment-based metrics, as well as its effectiveness in predicting downstream task performance. Comparative analyses against established metrics like BERTScore and ROUGE highlight the competitive performance of COSMIC.

Theory of mind (ToM) refers to humans’ ability to understand and infer the desires, beliefs, and intentions of others. The acquisition of ToM plays a key role in humans’ social cognition and interpersonal relations. Though indispensable for social intelligence, ToM is still lacking for modern AI and NLP systems since they cannot access the human mental state and cognitive process beneath the training corpus. To empower AI systems with the ToM ability and narrow the gap between them and humans, in this paper, we propose COKE: the first cognitive knowledge graph for machine theory of mind. Specifically, COKE formalizes ToM as a collection of 45k+ manually verified cognitive chains that characterize human mental activities and subsequent behavioral/affective responses when facing specific social circumstances. In addition, we further generalize COKE using LLMs and build a powerful generation model COLM tailored for cognitive reasoning. Experimental results in both automatic and human evaluation demonstrate the high quality of COKE, the superior ToM ability of COLM, and its potential to significantly enhance social applications.