ACL 2024 论文列表

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.

We identify two crucial limitations in the evaluation of recent parallel-integrated method Parallel Context Windows (PCW), which extends the maximum context lengths of language models, e.g., 2048 for LLaMA, by harnessing window-wise attention and positional embedding techniques. We first show that a simple yet strong baseline, weighted sum ensemble, is missing for the in-context few-shot classification. Moreover, on more challenging Chain-of-Thought (CoT) reasoning (e.g., HotpotQA), PCW would present unexpected deterioration regarding question miscomprehension and false inference. Based on our findings, we suggest that the existing PCW design may not guarantee sufficient improvement and practicality in handling lengthy documents in real-world applications. More community efforts on enabling language models’ long context understanding ability should be paid.

Numerous recent works aim to enhance the efficacy of Large Language Models (LLMs) through strategic prompting. In particular, the Optimization by PROmpting (OPRO) approach provides state-of-the-art performance by leveraging LLMs as optimizers where the optimization task is to find instructions that maximize the task accuracy. In this paper, we revisit OPRO for automated prompting with relatively small-scale LLMs, such as LLaMa-2 family and Mistral 7B. Our investigation reveals that OPRO shows limited effectiveness in small-scale LLMs, with limited inference capabilities constraining optimization ability. We suggest future automatic prompting engineering to consider both model capabilities and computational costs. Additionally, for small-scale LLMs, we recommend direct instructions that clearly outline objectives and methodologies as robust prompt baselines, ensuring efficient and effective prompt engineering in ongoing research.

Tabular data with text fields can be leveraged in applications such as financial risk assessment or medical diagnosis prediction. When employing multimodal approaches to make predictions based on these modalities, it is crucial to make the most appropriate modeling choices in terms of numerical feature encoding or fusion strategy. In this paper, we focus on multimodal classification tasks based on tabular datasets with text fields. We build on multimodal Transformers to propose the Tabular-Text Transformer (TTT), a tabular/text dual-stream Transformer network. This architecture includes a distance-to-quantile embedding scheme for numerical features and an overall attention module which concurrently considers self-attention and cross-modal attention. Further, we leverage the two well-informed modality streams to estimate whether a prediction is uncertain or not. To explain uncertainty in terms of feature values, we use a sampling-based approximation of Shapley values in a bimodal context, with two options for the value function. To show the efficacy and relevance of this approach, we compare it to six baselines and measure its ability to quantify and explain uncertainty against various methods. Our code is available at https://github.com/thomas-bonnier/TabularTextTransformer.

Knowledge distillation (KD) is a common approach to compress a teacher model to reduce its inference cost and memory footprint, by training a smaller student model. However, in the context of autoregressive language models (LMs), we empirically find that larger teacher LMs might dramatically result in a poorer student. In response to this problem, we conduct a series of analyses and reveal that different tokens have different teaching modes, neglecting which will lead to performance degradation. Motivated by this, we propose a simple yet effective adaptive teaching approach (ATKD) to improve the KD. The core of ATKD is to reduce rote learning and make teaching more diverse and flexible. Extensive experiments on 8 LM tasks show that, with the help of ATKD, various baseline KD methods can achieve consistent and significant performance gains (up to +3.04% average score) across all model types and sizes. More encouragingly, ATKD can improve the student model generalization effectively.

Speech-to-text (S2T) generation systems frequently face challenges in low-resource scenarios, primarily due to the lack of extensive labeled datasets. One emerging solution is constructing virtual training samples by interpolating inputs and labels, which has notably enhanced system generalization in other domains. Despite its potential, this technique’s application in S2T tasks has remained under-explored. In this paper, we delve into the utility of interpolation augmentation, guided by several pivotal questions. Our findings reveal that employing an appropriate strategy in interpolation augmentation significantly enhances performance across diverse tasks, architectures, and data scales, offering a promising avenue for more robust S2T systems in resource-constrained settings.

Large language models (LLMs) have shown an impressive ability to perform a wide range of tasks using in-context learning (ICL), where a few examples are used to describe a task to the model. However, the performance of ICL varies significantly with the choice of demonstrations, and previous research usually focuses on the data aspect ignoring the model’s effect. In this work, we first revisit the factors contributing to this variance from the model aspect, and find that the demonstration choice is both data- and model-dependent. We further propose a conjecture that the performance of a demonstration positively correlates with its contribution to the model’s understanding of the test samples, and accordingly propose a data- and model-dependent demonstration selection method, TopK + ConE. Empirically, our method yields consistent improvements in both language understanding and generation tasks with different model scales. Further analyses confirm that, besides the generality and stability under different circumstances, our method provides a unified explanation for the effectiveness of previous methods. Code is publicly available at https://github.com/Romainpkq/revisit_demon_selection_in_ICL.

This paper revisits recent code similarity evaluation metrics, particularly focusing on the application of Abstract Syntax Tree (AST) editing distance in diverse programming languages. In particular, we explore the usefulness of these metrics and compare them to traditional sequence similarity metrics. Our experiments showcase the effectiveness of AST editing distance in capturing intricate code structures, revealing a high correlation with established metrics. Furthermore, we explore the strengths and weaknesses of AST editing distance and prompt-based GPT similarity scores in comparison to BLEU score, execution match, and Jaccard Similarity. We propose, optimize, and publish an adaptable metric that demonstrates effectiveness across all tested languages, representing an enhanced version of Tree Similarity of Edit Distance (TSED).

Language Models (LMs) acquire parametric knowledge from their training process, embedding it within their weights. The increasing scalability of LMs, however, poses significant challenges for understanding a model’s inner workings and further for updating or correcting this embedded knowledge without the significant cost of retraining. This underscores the importance of unveiling exactly what knowledge is stored and its association with specific model components. Instance Attribution (IA) and Neuron Attribution (NA) offer insights into this training-acquired knowledge, though they have not been compared systematically. Our study introduces a novel evaluation framework to quantify and compare the knowledge revealed by IA and NA. To align the results of the methods we introduce the attribution method NA-Instances to apply NA for retrieving influential training instances, and IA-Neurons to discover important neurons of influential instances discovered by IA. We further propose a comprehensive list of faithfulness tests to evaluate the comprehensiveness and sufficiency of the explanations provided by both methods. Through extensive experiments and analysis, we demonstrate that NA generally reveals more diverse and comprehensive information regarding the LM’s parametric knowledge compared to IA. Nevertheless, IA provides unique and valuable insights into the LM’s parametric knowledge, which are not revealed by NA. Our findings further suggest the potential of a synergistic approach of combining the diverse findings of IA and NA for a more holistic understanding of an LM’s parametric knowledge.

Adaptive retrieval-augmented generation (ARAG) aims to dynamically determine the necessity of retrieval for queries instead of retrieving indiscriminately to enhance the efficiency and relevance of the sourced information. However, previous works largely overlook the evaluation of ARAG approaches, leading to their effectiveness being understudied. This work presents a benchmark, RetrievalQA, comprising 1,271 short-form questions covering new world and long-tail knowledge. The knowledge necessary to answer the questions is absent from LLMs; therefore, external information must be retrieved to answer correctly. This makes RetrievalQA a suitable testbed to evaluate existing ARAG methods. We observe that calibration-based methods heavily rely on threshold tuning, while vanilla prompting is inadequate for guiding LLMs to make reliable retrieval decisions. Based on our findings, we propose Time-Aware Adaptive Retrieval (TA-ARE), a simple yet effective method that helps LLMs assess the necessity of retrieval without calibration or additional training.

We propose a simple method that applies a large language model (LLM) to large-scale retrieval in zero-shot scenarios. Our method, the Large language model as Retriever (LameR), is built upon no other neural models but an LLM in a retrieval-augmented retrieval fashion, while breaking brute-force combinations of retrievers with LLMs and lifting the performance of zero-shot retrieval to be very competitive on benchmark datasets. Essentially, we propose to augment a query with its potential answers by prompting LLMs with a composition of the query and the query’s in-domain candidates. The candidates, regardless of correct or wrong, are obtained by a vanilla retrieval procedure on the target collection. As a part of the prompts, they are likely to help LLM generate more precise answers by pattern imitation or candidate summarization. Even if all the candidates are wrong, the prompts at least make LLM aware of in-collection patterns and genres. Moreover, due to the low performance of a self-supervised retriever, the LLM-based query augmentation becomes less effective as the retriever bottlenecks the whole pipeline. Therefore, we propose to leverage a non-parametric lexicon-based method (e.g., BM25) as the retrieval module to capture query-document overlap in a literal fashion. As such, LameR makes the retrieval procedure transparent to the LLM, thus circumventing the bottleneck.

Knowledge represented in Large Language Models (LLMs) is quite often incorrect and can also become obsolete over time. Updating knowledge via fine-tuning is computationally resource-hungry and not reliable, and so knowledge editing (KE) has developed as an effective and economical alternative to inject new knowledge or to fix factual errors in LLMs. Although there has been considerable interest in this area, current KE research exclusively focuses on monolingual settings, typically in English. However, what happens if the new knowledge is supplied in one language, but we would like to query an LLM in a different language? To address the problem of multilingual knowledge editing, we propose Retrieval-Augmented Multilingual Knowledge Editor (ReMaKE) to update knowledge in LLMs. ReMaKE can be used to perform model-agnostic knowledge editing in a multilingual setting. ReMaKE concatenates the new knowledge retrieved from a multilingual knowledge base with users’ prompts before querying an LLM. Our experimental results show that ReMaKE outperforms baseline knowledge editing methods by a significant margin and is scalable to real-word application scenarios. Our multilingual knowledge editing dataset (MzsRE) in 12 languages, the code, and additional project information are available at https://github.com/weixuan-wang123/ReMaKE.

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.

An essential requirement for a real-world Knowledge Base Question Answering (KBQA) system is the ability to detect the answerability of questions when generating logical forms. However, state-of-the-art KBQA models assume all questions to be answerable. Recent research has found that such models, when superficially adapted to detect answerability, struggle to satisfactorily identify the different categories of unanswerable questions, and simultaneously preserve good performance for answerable questions. Towards addressing this issue, we propose RetinaQA, a new KBQA model that unifies two key ideas in a single KBQA architecture: (a) discrimination over candidate logical forms, rather than generating these, for handling schema-related unanswerability, and (b) sketch-filling-based construction of candidate logical forms for handling data-related unaswerability. Our results show that RetinaQA significantly outperforms adaptations of state-of-the-art KBQA models in handling both answerable and unanswerable questions and demonstrates robustness across all categories of unanswerability. Notably, RetinaQA also sets a new state-of-the-art for answerable KBQA, surpassing existing models. We release our code base for further research: https://github.com/dair-iitd/RetinaQA.

Transformer-based methods have gone mainstream in multimodal sequential learning. The intra and inter modality interactions are captured by the query-key associations of multi-head attention. In this way, the calculated multimodal contexts (attentional results) are expected to be relevant to the query modality. However, in existing literature, the alignment degree between different calculated attentional results of the same query are under-explored. Based on this concern, we propose a new constrained scheme called Multimodal Contextual Contrast (MCC), which could align the multiple attentional results from both local and global perspectives, making the information capture more efficient. Concretely, the calculated attentional results of different modalities are mapped into a common feature space, those attentional vectors with the same query are considered as a positive group and the remaining sets are negative. From local perspective, we sample the negative groups for a positive group by randomly changing the sequential step of one specific context and keeping the other stay the same. From coarse global perspective, we divide all the contextual groups into two sets (i.e., aligned and unaligned), making the total score of aligned group relatively large. We extend the vectorial inner product operation for more input and calculate the aligned score for each multimodal group. Considering that the computational complexity scales exponentially to the number of modalities, we adopt stochastic expectation approximation (SEA) for the real process. The extensive experimental results on several tasks reveal the effectiveness of our contributions.

Recent progress in LLMs discussion suggests that multi-agent discussion improves the reasoning abilities of LLMs. In this work, we reevaluate this claim through systematic experiments, where we propose a novel group discussion framework to enrich the set of discussion mechanisms. Interestingly, our results show that a single-agent LLM with strong prompts can achieve almost the same best performance as the best existing discussion approach on a wide range of reasoning tasks and backbone LLMs. We observed that the multi-agent discussion performs better than a single agent only when there is no demonstration in the prompt. Further study reveals the common interaction mechanisms of LLMs during the discussion. Our code can be found in https://github.com/HKUST-KnowComp/LLM-discussion.

Task-oriented dialogue (TOD) systems are predominantly designed to be composed of several functional modules (e.g. dialogue state tracker, dialogue policy, natural language generation) whether they are pipeline or end-to-end architectures. However, this modular design not only heavily relies on massive fully-annotated data, but also suffers from many intrinsic drawbacks, such as serious error accumulation, poor generalization ability, high customization cost, and low fault tolerance rate. In this paper, we rethink the architecture of the task-oriented dialogue systems and propose a novel fully zero-shot autonomous TOD agent, named AutoTOD, where all the delicate modules in traditional TOD systems are deprecated and all it needs is a general-purpose instruction-following language model (e.g. GPT-4). AutoTOD only leverages a simple instruction schema consisting of the description of tasks and external APIs, and can autonomously decide what to do at each dialogue turn, including asking for information, calling APIs, summarizing API results, and correcting previous mistakes. Moreover, we propose a simulation-based evaluation framework to better validate the abilities of TOD models in real-life scenarios. Extensive experiments conducted on the MultiWOZ and SGD datasets show the superior task completion ability and flexible language skills of AutoTOD.

Large Language Models (LLMs) have demonstrated impressive capabilities for generalizing in unseen tasks. In the Named Entity Recognition (NER) task, recent advancements have seen the remarkable improvement of LLMs in a broad range of entity domains via instruction tuning, by adopting entity-centric schema. In this work, we explore the potential enhancement of the existing methods by incorporating negative instances into training. Our experiments reveal that negative instances contribute to remarkable improvements by (1) introducing contextual information, and (2) clearly delineating label boundaries. Furthermore, we introduce an efficient longest common subsequence (LCS) matching algorithm, which is tailored to transform unstructured predictions into structured entities. By integrating these components, we present GNER, a Generative NER system that shows improved zero-shot performance across unseen entity domains. Our comprehensive evaluation illustrates our system’s superiority, surpassing state-of-the-art (SoTA) methods by 9 F1 score in zero-shot evaluation.

Evaluating multilingual summarization evaluation metrics, i.e., meta-evaluation, is challenging because of the difficulty of human annotation collection. Therefore, we investigate an efficient multilingual meta-evaluation framework that uses machine translation systems to transform a monolingual meta-evaluation dataset into multilingual versions. To this end, we introduce a statistical test to verify the transformed dataset quality by checking the meta-evaluation result consistency on the original dataset and back-translated dataset. With this quality verification method, we transform an existing English summarization meta-evaluation dataset, RoSE, into 30 languages, and conduct a multilingual meta-evaluation of several representative automatic evaluation metrics. In our meta-evaluation, we find that metric performance varies in different languages and neural metrics generally outperform classical text-matching-based metrics in non-English languages. Moreover, we identify a two-stage evaluation method with superior performance, which first translates multilingual texts into English and then performs evaluation. We make the transformed datasets publicly available to facilitate future research.