NAACL 2024 论文列表

While large language models (LMs) demonstrate remarkable performance, they encounter challenges in providing accurate responses when queried for information beyond their pre-trained memorization. Although augmenting them with relevant external information can mitigate these issues, failure to consider the necessity of retrieval may adversely affect overall performance. Previous research has primarily focused on examining how entities influence retrieval models and knowledge recall in LMs, leaving other aspects relatively unexplored. In this work, our goal is to offer a more detailed, fact-centric analysis by exploring the effects of combinations of entities and relations. To facilitate this, we construct a new question answering (QA) dataset called WiTQA (Wikipedia Triple Question Answers). This dataset includes questions about entities and relations of various popularity levels, each accompanied by a supporting passage. Our extensive experiments with diverse LMs and retrievers reveal when retrieval does not consistently enhance LMs from the viewpoints of fact-centric popularity. Confirming earlier findings, we observe that larger LMs excel in recalling popular facts. However, they notably encounter difficulty with infrequent entity-relation pairs compared to retrievers. Interestingly, they can effectively retain popular relations of less common entities. We demonstrate the efficacy of our finer-grained metric and insights through an adaptive retrieval system that selectively employs retrieval and recall based on the frequencies of entities and relations in the question.

Large Language Models (LLMs) have shown to be capable of various tasks, yet their capability in interpreting and reasoning over tabular data remains an underexplored area. In this context, this study investigates from three core perspectives: the robustness of LLMs to structural perturbations in tables, the comparative analysis of textual and symbolic reasoning on tables, and the potential of boosting model performance through the aggregation of multiple reasoning pathways. We discover that structural variance of tables presenting the same content reveals a notable performance decline, particularly in symbolic reasoning tasks. This prompts the proposal of a method for table structure normalization. Moreover, textual reasoning slightly edges out symbolic reasoning, and a detailed error analysis reveals that each exhibits different strengths depending on the specific tasks. Notably, the aggregation of textual and symbolic reasoning pathways, bolstered by a mix self-consistency mechanism, resulted in achieving SOTA performance, with an accuracy of 73.6% on WikiTableQuestions, representing a substantial advancement over previous existing table processing paradigms of LLMs.

Making moral judgments is an essential step toward developing ethical AI systems. Prevalent approaches are mostly implemented in a bottom-up manner, which uses a large set of annotated data to train models based on crowd-sourced opinions about morality. These approaches have been criticized for potentially overgeneralizing a limited group of annotators’ moral stances and lacking explainability. This work proposes a flexible top-down framework to steer (Large) Language Models to perform moral reasoning with well-established moral theories from interdisciplinary research. The theory-guided top-down framework can incorporate various moral theories. Our experiments demonstrate the effectiveness of the proposed framework on datasets derived from moral theories. Furthermore, we show the alignment between different moral theories and existing morality datasets. Our analysis exhibits the potential and flaws in existing resources (models and datasets) in developing explainable moral judgment-making systems.

As large language models (LLMs) have increased in their capabilities, so doestheir potential for dual use. To reduce harmful outputs, produces and vendors ofLLMs have used reinforcement learning with human feedback (RLHF). In tandem,LLM vendors have been increasingly enabling fine-tuning of their most powerfulmodels. However, concurrent work has shown that fine-tuning can remove RLHFprotections. We may expect that the most powerful models currently available(GPT-4) are less susceptible to fine-tuning attacks. In this work, we show the contrary: fine-tuning allows attackers to remove RLHFprotections with as few as 340 examples and a 95% success rate. These trainingexamples can be automatically generated with weaker models. We further show thatremoving RLHF protections does not decrease usefulness on non-censored outputs,providing evidence that our fine-tuning strategy does not decrease usefulnessdespite using weaker models to generate training data. Our results show the needfor further research on protections on LLMs.

Evaluating the reliability of news sources is a routine task for journalists and organizations committed to acquiring and disseminating accurate information.Recent research has shown that predicting sources’ reliability represents an important first-prior step in addressing additional challenges such as fake news detection and fact-checking.In this paper, we introduce a novel approach for source reliability estimation that leverages reinforcement learning strategies for estimating the reliability degree of news sources. Contrary to previous research, our proposed approach models the problem as the estimation of a reliability degree, and not a reliability label, based on how all the news media sources interact with each other on the Web.We validated the effectiveness of our method on a news media reliability dataset that is an order of magnitude larger than comparable existing datasets. Results show that the estimated reliability degrees strongly correlates with journalists-provided scores (Spearman=0.80) and can effectively predict reliability labels (macro-avg. F1 score=81.05).We release our implementation and dataset, aiming to provide a valuable resource for the NLP community working on information verification.

To meet the requirements of real-world applications, it is essential to control generations of large language models (LLMs). Prior research has tried to introduce reinforcement learning (RL) into controllable text generation while most existing methods suffer from overfitting issues (finetuning-based methods) or semantic collapse (post-processing methods). However, current RL methods are generally guided by coarse-grained (sentence/paragraph-level) feedback, which may lead to suboptimal performance owing to semantic twists or progressions within sentences. To tackle that, we propose a novel reinforcement learning algorithm named TOLE which formulates TOken-LEvel rewards for controllable text generation, and employs a “first-quantize-then-noise” paradigm to enhance the robustness of the RL algorithm. Furthermore, TOLE can be flexibly extended to multiple constraints with little computational expense. Experimental results show that our algorithm can achieve superior performance on both single-attribute and multi-attribute control tasks. We have released our codes at https://github.com/WindyLee0822/CTG.

Language usage is related to speaker age, gender, moral concerns, political ideology, and other attributes. Current state-of-the-art methods for predicting these attributes take a speaker’s utterances as input and provide a prediction per speaker attribute. Most of these approaches struggle to handle a large number of utterances per speaker. This difficulty is primarily due to the computational constraints of the models. Additionally, only a subset of speaker utterances may be relevant to specific attributes. In this paper, we formulate speaker attribute prediction as a Multiple Instance Learning (MIL) problem and propose RL-MIL, a novel approach based on Reinforcement Learning (RL) that effectively addresses both of these challenges. Our experiments demonstrate that our RL-based methodology consistently outperforms previous approaches across a range of related tasks: predicting speakers’ psychographics and demographics from social media posts, and political ideologies from transcribed speeches. We create synthetic datasets and investigate the behavior of RL-MIL systematically. Our results show the success of RL-MIL in improving speaker attribute prediction by learning to select relevant speaker utterances.

Continual learning aims at incrementally acquiring new knowledge while not forgetting existing knowledge. To overcome catastrophic forgetting, methods are either rehearsal-based, i.e., store data examples from previous tasks for data replay, or isolate parameters dedicated to each task. However, rehearsal-based methods raise privacy and memory issues, and parameter-isolation continual learning does not consider interaction between tasks, thus hindering knowledge transfer. In this work, we propose MoCL, a rehearsal-free **Mo**dular and **C**ompositional Continual **L**earning framework which continually adds new modules to language models and composes them with existing modules. Experiments on various benchmarks show that MoCL outperforms state of the art and effectively facilitates knowledge transfer.

We present a game-theoretic model of pragmatics that we call ReCo (for Regularized Conventions). This model formulates pragmatic communication as a game in which players are rewarded for communicating successfully and penalized for deviating from a shared, “default” semantics. As a result, players assign utterances context-dependent meanings that jointly optimize communicative success and naturalness with respect to speakers’ and listeners’ background knowledge of language. By using established game-theoretic tools to compute equilibrium strategies for this game, we obtain principled pragmatic language generation procedures with formal guarantees of communicative success. Across several datasets capturing real and idealized human judgments about pragmatic implicature, ReCo matches, or slightly improves upon, predictions made by Iterated Best Response and Rational Speech Acts models of language understanding.

A current limitation of Generative AI (GenAI) is its propensity to hallucinate. While Large Language Models (LLM) have taken the world by storm, without eliminating or at least reducing hallucination, real-world GenAI systems will likely continue to face challenges in user adoption. In the process of deploying an enterprise application that produces workflows from natural language requirements, we devised a system leveraging Retrieval-Augmented Generation (RAG) to improve the quality of the structured output that represents such workflows. Thanks to our implementation of RAG, our proposed system significantly reduces hallucination and allows the generalization of our LLM to out-of-domain settings. In addition, we show that using a small, well-trained retriever can reduce the size of the accompanying LLM at no loss in performance, thereby making deployments of LLM-based systems less resource-intensive.

The recent progress of AI can be largely attributed to large language models (LLMs). However, their escalating memory requirements introduce challenges for machine learning (ML) researchers and engineers. Addressing this requires developers to partition a large model to distribute it across multiple GPUs or TPUs. This necessitates considerable coding and intricate configuration efforts with existing model parallel tools, such as Megatron-LM, DeepSpeed, and Alpa. These tools require users’ expertise in machine learning systems (MLSys), creating a bottleneck in LLM development, particularly for developers without MLSys background. In this work, we present RedCoast (Redco), a lightweight and user-friendly tool crafted to automate distributed training and inference for LLMs, as well as to simplify ML pipeline development. The design of Redco emphasizes two key aspects. Firstly, to automate model parallelism, our study identifies two straightforward rules to generate tensor parallel strategies for any given LLM. Integrating these rules into Redco facilitates effortless distributed LLM training and inference, eliminating the need of additional coding or complex configurations. We demonstrate the effectiveness by applying Redco on a set of LLM architectures, such as GPT-J, LLaMA, T5, and OPT, up to the size of 66B. Secondly, we propose a mechanism that allows for the customization of diverse ML pipelines through the definition of merely three functions, avoiding redundant and formulaic code like multi-host related processing. This mechanism proves adaptable across a spectrum of ML algorithms, from foundational language modeling to complex algorithms like meta-learning and reinforcement learning. As a result, Redco implementations exhibit significantly fewer lines of code compared to their official counterparts. RedCoast (Redco) has been released under Apache 2.0 license at https://github.com/tanyuqian/redco.

Large language models (LLMs) are able to solve various tasks with only a few demonstrations utilizing their in-context learning (ICL) abilities.However, LLMs often rely on their pre-trained semantic priors of demonstrations rather than on the input-label relationships to proceed with ICL prediction. In this work, we term this phenomenon as the ‘Demonstration Shortcut’.While previous works have primarily focused on improving ICL prediction results for predefined tasks, we aim to rectify the Demonstration Shortcut, thereby enabling the LLM to effectively learn new input-label relationships from demonstrations.To achieve this, we introduce In-Context Calibration, a demonstration-aware calibration method.We evaluate the effectiveness of the proposed method in two settings: (1) the Original ICL Task using the standard label space and (2) the Task Learning setting, where the label space is replaced with semantically unrelated tokens.In both settings, In-Context Calibration demonstrates substantial improvements, with results generalized across three LLM families (OPT, GPT, and Llama2) under various configurations.

While the recommendation system (RS) has advanced significantly through deep learning, current RS approaches usually train and fine-tune models on task-specific datasets, limiting their generalizability to new recommendation tasks and their ability to leverage external knowledge due to model scale and data size constraints. Thus, we designed an LLM-powered autonomous recommender agent, RecMind, which is capable of leveraging external knowledge, utilizing tools with careful planning to provide zero-shot personalized recommendations. We propose a Self-Inspiring algorithm to improve the planning ability. At each intermediate step, the LLM “self-inspires” to consider all previously explored states to plan for the next step. This mechanism greatly improves the model’s ability to comprehend and utilize historical information in planning for recommendation. We evaluate RecMind’s performance in various recommendation scenarios. Our experiment shows that RecMind outperforms existing zero/few-shot LLM-based recommendation baseline methods in various tasks and achieves comparable performance to a fully trained recommendation model P5.

The impressive performance of recent language models across a wide range of tasks suggests that they possess a degree of abstract reasoning skills. Are these skills general and transferable, or specialized to specific tasks seen during pretraining? To disentangle these effects, we propose an evaluation framework based on “counterfactual” task variants that deviate from the default assumptions underlying standard tasks. Across a suite of 11 tasks, we observe nontrivial performance on the counterfactual variants, but nevertheless find that performance substantially and consistently degrades compared to the default conditions. This suggests that while current LMs may possess abstract task-solving skills to an extent, they often also rely on narrow, non-transferable procedures for task-solving. These results motivate a more careful interpretation of language model performance that teases apart these aspects.

Large Language Models (LLMs) have demonstrated good performance in many reasoning tasks, but they still struggle with some complicated reasoning tasks including logical reasoning. One non-negligible reason for LLMs’ suboptimal performance on logical reasoning is their overlooking of understanding logical fallacies correctly. To evaluate LLMs’ capability of logical fallacy understanding (LFU), we propose five concrete tasks from three cognitive dimensions of WHAT, WHY, and HOW in this paper. Towards these LFU tasks, we have successfully constructed a new dataset LFUD based on GPT-4 accompanied by a little human effort. Our extensive experiments justify that our LFUD can be used not only to evaluate LLMs’ LFU capability, but also to fine-tune LLMs to obtain significantly enhanced performance on logical reasoning.

While text summarization is a well-known NLP task, in this paper, we introduce a novel and useful variant of it called functionality extraction from Git README files. Though this task is a text2text generation at an abstract level, it involves its own peculiarities and challenges making existing text2text generation systems not very useful. The motivation behind this task stems from a recent surge in research and development activities around the use of large language models for code-related tasks, such as code refactoring, code summarization, etc. We also release a human-annotated dataset called FuncRead, and develop a battery of models for the task. Our exhaustive experimentation shows that small size fine-tuned models beat any baseline models that can be designed using popular black-box or white-box large language models (LLMs) such as ChatGPT and Bard. Our best fine-tuned 7 Billion CodeLlama model exhibit 70% and 20% gain on the F1 score against ChatGPT and Bard respectively.

Current logical reasoning evaluations of Large Language Models (LLMs) primarily focus on single-turn and static environments, such as arithmetic problems. The crucial problem of multi-turn, strategic reasoning is under-explored. In this work, we analyze the multi-turn strategic reasoning of LLMs through text-driven complete- and incomplete-information gaming, e.g., board games (Tic-Tac-Toe, Connect-4) and poker games (Texas Hold’em Poker). Specifically, we consider two distinct scenarios: 1) Online Racing, featuring multiple LLMs/agents to facilitate direct competition and comparison; 2) Offline Probing, constructing targeted questions with verified ground truth to evaluate LLMs’ strategic behaviors. Experimental results demonstrate that existing state-of-the-art LLMs and reasoning schemes are largely ineffective for strategic reasoning tasks. To mitigate these limitations, we propose a simple yet effective Recursively Thinking-Ahead (ReTA) agent, incorporating a recursive prompting mechanism that automatically analyzes the opponents’ future moves/actions and assigns reward signals for these situations, to strengthen the strategic reasoning of LLMs. We hope our work could spur further research and exploration in the multi-turn strategic reasoning of LLMs. The code is available at https://github.com/jinhaoduan/ReTA.

Our world is marked by unprecedented technological, global, and socio-political transformations, posing a significant challenge to textto-image generative models. These models encode factual associations within their parameters that can quickly become outdated, diminishing their utility for end-users. To that end, we introduce ReFACT, a novel approach for editing factual associations in text-to-image models without relaying on explicit input from end-users or costly re-training. ReFACT updates the weights of a specific layer in the text encoder, modifying only a tiny portion of the model’s parameters and leaving the rest of the model unaffected.We empirically evaluate ReFACT on an existing benchmark, alongside a newly curated dataset.Compared to other methods, ReFACT achieves superior performance in both generalization to related concepts and preservation of unrelated concepts.Furthermore, ReFACT maintains image generation quality, making it a practical tool for updating and correcting factual information in text-to-image models.

Despite remarkable advancements in mitigating hallucinations in large language models (LLMs) by retrieval augmentation, it remains challenging to measure the reliability of LLMs using static question-answering (QA) data. Specifically, given the potential of data contamination (e.g., leading to memorization), good static benchmark performance does not ensure that model can reliably use the provided evidence for responding, which is essential to avoid hallucination when the required knowledge is new or private. Inspired by adversarial machine learning, we investigate the feasibility of automatically perturbing existing static one for dynamic evaluation. Specifically, this paper presents ReEval, an LLM-based framework using prompt chaining to perturb the original evidence for generating new test cases for evaluating the LLMs’ reliability in using new evidence for answering.We implement ReEval using ChatGPT and evaluate the resulting variants of two popular open-domain QA datasets on a collection ofLLMs under various prompting settings. Our generated data is human-readable and useful to trigger hallucination in LLM. Accurate models on static data are observed to produce unsupported answers from the perturbed evidence, with pronounced accuracy drops across LLMs including GPT-4. We find that our adversarial examples are transferable across all considered LLMs. The examples generated by a small model can be used to evaluate a much larger model, making our approach cost-effective.