NAACL 2025 Papers

Trustworthiness in model predictions is crucial for safety-critical applications in the real world. However, deep neural networks often suffer from the issues of uncertainty estimation, such as miscalibration. In this study, we propose k-Nearest Neighbor Uncertainty Estimation (kNN-UE), which is a new uncertainty estimation method that uses not only the distances from the neighbors, but also the ratio of labels in the neighbors. Experiments on sentiment analysis, natural language inference, and named entity recognition show that our proposed method outperforms the baselines and recent density-based methods in several calibration and uncertainty metrics. Moreover, our analyses indicate that approximate nearest neighbor search techniques reduce the inference overhead without significantly degrading the uncertainty estimation performance when they are appropriately combined.

Complex table question answering (TQA) aims to answer questions that require complex reasoning, such as multi-step or multi-category reasoning, over data represented in tabular form. Previous approaches demonstrate notable performance by leveraging either closed-source large language models (LLMs) or fine-tuned open-weight LLMs. However, fine-tuning LLMs requires high-quality training data, which is costly to obtain. The use of closed-source LLMs poses accessibility challenges and leads to reproducibility issues. In this paper, we propose Multi Agent Collaboration with Tool use (MACT), a framework that requires neither fine-tuning nor closed-source models. In MACT, a planning agent and a coding agent that also make use of tools collaborate for TQA. MACT outperforms previous SoTA systems on three out of four benchmarks and performs comparably to the larger and more expensive closed-source model GPT-4 on two benchmarks, even when using only open-weight models without any fine-tuning. Our extensive analyses prove the effectiveness of MACT’s multi-agent collaboration in TQA. We release our code publicly.

Open-source large language models (Os-LLMs) propel the democratization of natural language research by giving the flexibility to augment or update model parameters for performance improvement. Nevertheless, like proprietary LLMs, Os-LLMs offer poorer performance on low-resource languages (LRLs) than high-resource languages (HRLs), owing to smaller amounts of training data and underrepresented vocabulary. On the other hand, continual pre-training (CPT) with large amounts of language-specific data is a costly proposition in terms of data acquisition and computational resources. Our goal is to drastically reduce CPT cost.To that end, we first develop a new algorithm to select a subset of texts from a larger corpus. We show the effectiveness of our technique using very little CPT data. In search of further improvement, we design a new algorithm to select tokens to include in the LLM vocabulary.We experiment with the recent Llama-3 model and nine Indian languages with diverse scripts and extent of resource availability.For evaluation, we use IndicGenBench, a generation task benchmark dataset for Indic languages. We experiment with various CPT corpora and augmented vocabulary size and offer insights across language families.

Misinformation spreads rapidly on social media, confusing the truth and targeting potentially vulnerable people. To effectively mitigate the negative impact of misinformation, it must first be accurately detected before applying a mitigation strategy, such as X’s community notes, which is currently a manual process. This study takes a knowledge-based approach to misinformation detection, modelling the problem similarly to one of natural language inference. The EffiARA annotation framework is introduced, aiming to utilise inter- and intra-annotator agreement to understand the reliability of each annotator and influence the training of large language models for classification based on annotator reliability. In assessing the EffiARA annotation framework, the Russo-Ukrainian Conflict Knowledge-Based Misinformation Classification Dataset (RUC-MCD) was developed and made publicly available. This study finds that sample weighting using annotator reliability performs the best, utilising both inter- and intra-annotator agreement and soft label training. The highest classification performance achieved using Llama-3.2-1B was a macro-F1 of 0.757 and 0.740 using TwHIN-BERT-large.

Large Language Models (LLMs) have demonstrated impressive performance on a wide range of natural language processing (NLP) tasks, primarily through in-context learning (ICL). In ICL, the LLM is provided with examples that represent a given task such that it learns to generate answers for test inputs. However, access to these in-context examples is not guaranteed especially for low-resource or massively multilingual tasks. In this work, we propose an unsupervised approach to mine in-context examples for machine translation (MT), enabling unsupervised MT (UMT) across different languages. Our approach begins with word-level mining to acquire word translations that are then used to perform sentence-level mining. As the quality of mined parallel pairs may not be optimal due to noise or mistakes, we introduce a filtering criterion to select the optimal in-context examples from a pool of unsupervised parallel sentences. We evaluate our approach using two multilingual LLMs on 288 directions from the FLORES-200 dataset (CITATION) and analyze the impact of various linguistic features on performance. Our findings demonstrate the effectiveness of our unsupervised approach in mining in-context examples for MT, leading to better or comparable translation performance as translation with regular in-context samples (extracted from human-annotated data), while also outperforming the other state-of-the-art UMT methods by an average of 7 BLEU points.

Hate speech (HS) erodes the inclusiveness of online users and propagates negativity and division. Counterspeech has been recognized as a way to mitigate the harmful consequences. While some research has investigated the impact of user-generated counterspeech on social media platforms, few have examined and modeled haters’ reactions toward counterspeech, despite the immediate alteration of haters’ attitudes being an important aspect of counterspeech. This study fills the gap by analyzing the impact of counterspeech from the hater’s perspective, focusing on whether the counterspeech leads the hater to reenter the conversation and if the reentry is hateful. We compile the Reddit Echoes of Hate dataset (ReEco), which consists of triple-turn conversations featuring haters’ reactions, to assess the impact of counterspeech. To predict haters’ behaviors, we employ two strategies: a two-stage reaction predictor and a three-way classifier. The linguistic analysis sheds insights on the language of counterspeech to hate eliciting different haters’ reactions. Experimental results demonstrate that the 3-way classification model outperforms the two-stage reaction predictor, which first predicts reentry and then determines the reentry type. We conclude the study with an assessment showing the most common errors identified by the best-performing model.

Recent advancements in large language models (LLM) capable of processing extremely long texts highlight the need for a dedicated evaluation benchmark to assess their long-context capabilities. However, existing methods, like the needle-in-a-haystack test, do not effectively assess whether these models fully utilize contextual information, raising concerns about the reliability of current evaluation techniques. To thoroughly examine the effectiveness of existing benchmarks, we introduce a new metric called information coverage (IC), which quantifies the proportion of the input context necessary for answering queries. Our findings indicate that current benchmarks exhibit low IC; although the input context may be extensive, the actual usable context is often limited. To address this, we present ETHIC, a novel benchmark designed to assess LLMs’ ability to leverage the entire context. Our benchmark comprises 1,986 test instances spanning four long-context tasks with high IC scores in the domains of books, debates, medicine, and law. Our evaluations reveal significant performance drops in contemporary LLMs, highlighting a critical challenge in managing long contexts. Our benchmark is available at https://github.com/dmis-lab/ETHIC.

We present ESPnet-SpeechLM, an open toolkit designed to democratize the development of speech language models (SpeechLMs) and voice-driven agentic applications. The toolkit standardizes speech processing tasks by framing them as universal sequential modeling problems, encompassing a cohesive workflow of data preprocessing, pre-training, inference, and task evaluation. With ESPnet-SpeechLM, users can easily define task templates and configure key settings, enabling seamless and streamlined SpeechLM development. The toolkit ensures flexibility, efficiency, and scalability by offering highly configurable modules for every stage of the workflow. To illustrate its capabilities, we provide multiple use cases demonstrating how competitive SpeechLMs can be constructed with ESPnet-SpeechLM, including a 1.7B-parameter model pre-trained on both text and speech tasks, across diverse benchmarks. The toolkit and its recipes are fully transparent and reproducible at: https://github.com/espnet/espnet/tree/speechlm.

Advancements in audio foundation models (FMs) have fueled interest in end-to-end (E2E) spoken dialogue systems, but different web interfaces for each system makes it challenging to compare and contrast them effectively. Motivated by this, we introduce an open-source, user-friendly toolkit designed to build unified web interfaces for various cascaded and E2E spoken dialogue systems. Our demo further provides users with the option to get on-the-fly automated evaluation metrics such as (1) latency, (2) ability to understand user input, (3) coherence, diversity, and relevance of system response, and (4) intelligibility and audio quality of system output. Using the evaluation metrics, we compare various cascaded and E2E spoken dialogue systems with a human-human conversation dataset as a proxy. Our analysis demonstrates that the toolkit allows researchers to effortlessly compare and contrast different technologies, providing valuable insights such as current E2E systems having poorer audio quality and less diverse responses. An example demo produced using our toolkit is publicly available here: https://huggingface.co/spaces/Siddhant/Voice_Assistant_Demo.

In languages without orthographic word boundaries, NLP models perform _word segmentation_, either as an explicit preprocessing step or as an implicit step in an end-to-end computation. This paper shows that Chinese NLP models are vulnerable to _morphological garden path errors_—errors caused by a failure to resolve local word segmentation ambiguities using sentence-level morphosyntactic context. We propose a benchmark, _ERAS_, that tests a model’s vulnerability to morphological garden path errors by comparing its behavior on sentences with and without local segmentation ambiguities. Using ERAS, we show that word segmentation models make morphological garden path errors on locally ambiguous sentences, but do not make equivalent errors on unambiguous sentences. We further show that sentiment analysis models with character-level tokenization make implicit garden path errors, even without an explicit word segmentation step in the pipeline. Our results indicate that models’ segmentation of Chinese text often fails to account for morphosyntactic context.

Speculative decoding emerges as a pivotal technique for enhancing the inference speed of Large Language Models (LLMs). Despite recent research aiming to improve prediction efficiency, multi-sample speculative decoding has been overlooked due to varying numbers of accepted tokens within a batch in the verification phase. Vanilla method adds padding tokens in order to ensure that the number of new tokens remains consistent across samples. However, this increases the computational and memory access overhead, thereby reducing the speedup ratio. We propose a novel method that can resolve the issue of inconsistent tokens accepted by different samples without necessitating an increase in memory or computing overhead. Furthermore, our proposed method can handle the situation where the prediction tokens of different samples are inconsistent without the need to add padding tokens. Sufficient experiments demonstrate the efficacy of our method. Our code will be released later.

There has been a rising interest in utilizing tools in applications of autonomous agents based on large language models (LLMs) to address intricate real-world tasks. To develop LLMbased agents, it usually requires LLMs to understand many tool functions from different tool documentations. However, these documentations could be diverse, redundant, or incomplete, which immensely affects the capability of LLMs in using tools. Current LLMs exhibit satisfactory instruction-following capabilities based on instruction-following fine-tuning process. Motivated by this, in this paper, we introduce EASYTOOL, a framework transforming diverse and lengthy tool documentation into a unified and concise tool instruction to fully leverage instruction-following capabilities of LLMs for easier tool usage. EASYTOOL purifies essential information from extensive tool documentation of different sources, and elaborates a unified interface (i.e., tool instruction) to offer standardized tool descriptions and functionalities for LLM-based agents. Extensive experiments on multiple different tasks demonstrate that EASYTOOL can significantly reduce token consumption and improve the performance of LLM-based agents on tool utilization in real-world scenarios. Our code is available in supplemental materials. Our code is available at https://github.com/microsoft/JARVIS/tree/main/easytool.

Vector representations have been pivotal in advancing natural language processing (NLP), with prior research focusing on embedding techniques for mathematical expressions using mathematically equivalent formulations. While effective, these approaches are constrained by the size and diversity of training data. In this work, we address these limitations by introducing E-Gen, a novel e-graph-based dataset generation scheme that synthesizes large and diverse mathematical expression datasets, surpassing prior methods in size and operator variety. Leveraging this dataset, we train embedding models using two strategies: (1) generating mathematically equivalent expressions, and (2) contrastive learning to explicitly group equivalent expressions. We evaluate these embeddings on both in-distribution and out-of-distribution mathematical language processing tasks, comparing them against prior methods. Finally, we demonstrate that our embedding-based approach outperforms state-of-the-art large language models (LLMs) on several tasks, underscoring the necessity of optimizing embedding methods for the mathematical data modality. The source code and datasets are available at https://github.com/MLPgroup/E-Gen.

Large language models (LLMs) can learn vast amounts of knowledge from diverse domains during pre-training. However, long-tail knowledge from specialized domains is often scarce and underrepresented, rarely appearing in the models’ memorization. Prior work has shown that in-context learning (ICL) with retriever augmentation can help LLMs better capture long-tail knowledge, reducing their reliance on pre-trained data. Despite these advances, we observe that LLM predictions for long-tail questions remain uncertain to variations in retrieved samples. To take advantage of the uncertainty in ICL for guiding LLM predictions toward correct answers on long-tail samples, we propose a reinforcement learning-based dynamic uncertainty ranking method for retrieval-augmented ICL that accounts for the varying impact of each retrieved sample on LLM predictions. Our approach prioritizes more informative and stable samples while demoting misleading ones, updating rankings based on the feedback from the LLM w.r.t. each retrieved sample. To enhance training efficiency and reduce query costs, we introduce a learnable dynamic ranking threshold, adjusted when the model encounters negative prediction shifts. Experimental results on various question-answering datasets from different domains show that our method outperforms the best baseline by 2.76%, with a notable 5.96% boost in accuracy on long-tail questions that elude zero-shot inference. Our code is available at https://github.com/Yu-shuyan/uncertian_ranker.

Research has shown an effectiveness of reasoning (e.g. Chain-of-Thought), planning (e.g. SelfAsk) and retrieval augmented generation strategies to improve performance of Large Language Models (LLMs) on various tasks, such as question answering. However, using a single fixed strategy for answering all different kinds of questions is sub-optimal in performance and inefficient in terms of generated tokens and retrievals. In our work, we propose a novel technique, DyPlan, to induce a dynamic strategy selection process in LLMs for cost-effective question-answering. DyPlan incorporates an initial decision step to select the most suitable strategy conditioned on the input question and guides the LLM’s response generation accordingly. We extend DyPlan to DyPlan-verify, adding an internal verification and correction process to further enrich the generated answer. Experimentation on three prominent multi-hop question answering (MHQA) datasets reveals how DyPlan can improve model performance by 7-13% while reducing the cost by 11-32% relative to the best baseline model.

With the extensive deployment of Large Language Models (LLMs), ensuring their safety has become increasingly critical. However, existing defense methods often struggle with two key issues: (i) inadequate defense capabilities, particularly in domain-specific scenarios like chemistry, where a lack of specialized knowledge can lead to the generation of harmful responses to malicious queries. (ii) over-defensiveness, which compromises the general utility and responsiveness of LLMs. To mitigate these issues, we introduce a multi-agents-based defense framework, Guide for Defense (G4D), which leverages accurate external information to provide an unbiased summary of user intentions and analytically grounded safety response guidance. Extensive experiments on popular jailbreak attacks and benign datasets show that our G4D can enhance LLM’s robustness against jailbreak attacks on general and domain-specific scenarios without compromising the model’s general functionality.

The fine-tuning of pre-trained language models has resulted in the widespread availability of task-specific models. Model merging offers an efficient way to create multi-task models by combining these fine-tuned models at the parameter level, without the need for training data or joint training on multiple datasets. Existing merging approaches typically involve scaling the parameters model-wise or integrating parameter importance parameter-wise. Both approaches exhibit their own weaknesses, leading to a notable performance gap compared to multi-task fine-tuning. In this paper, we unify these seemingly distinct strategies into a more general merging framework, and introduce Dynamic Fisher-weighted Merging (DF-Merge). Specifically, candidate models are associated with a set of coefficients that linearly scale their fine-tuned parameters. Bayesian optimization is applied to dynamically adjust these coefficients, aiming to maximize overall performance on validation sets. Each iteration of this process integrates parameter importance based on the Fisher information conditioned by the coefficients. Experimental results show that DF-Merge outperforms strong baselines across models of different sizes and a variety of tasks. Our analysis shows that the effectiveness of DF-Merge arises from the unified view of merging and that near-optimal performance is achievable in a few iterations, even with minimal validation data.

This paper presents an efficient dual-stream early fusion method for sign language translation. Inspired by the brain’s ability to process color, shape, and motion simultaneously, the method explores complex dependencies between RGB and keypoint streams, improving speed and efficiency. A key challenge is extracting complementary features from both streams while ensuring global semantic consistency to avoid conflicts and improve generalization. To address this issue, we propose a hypernetwork-based fusion strategy that effectively extracts salient features from RGB and keypoint streams, alongside a partial shortcut connection training method to strengthen the complementary information between the dual streams. Additionally, we introduce self-distillation and SST contrastive learning to maintain feature advantages while aligning the global semantic space. Experiments show that our method achieves state-of-the-art performance on two public sign language datasets, reducing model parameters by about two-thirds.

Mixture-of-Experts (MoE) models have shown remarkable capability in instruction tuning, especially when the number of tasks scales. However, previous methods simply merge all training tasks (e.g. creative writing, coding, and mathematics) and apply fixed sampling weights, without considering the importance of different tasks as the model training state changes. In this way, the most helpful data cannot be effectively distinguished, leading to suboptimal model performance. To reduce the potential redundancies of datasets, we make the first attempt and propose a novel dynamic data mixture for MoE instruction tuning. Specifically, inspired by MoE’s token routing preference, we build dataset-level representations and then capture the subtle differences among datasets. Finally, we propose to dynamically adjust the sampling weight of datasets by their inter-redundancies, thus maximizing global performance under a limited training budget. The experimental results on two MoE models demonstrate the effectiveness of our approach on both downstream knowledge & reasoning tasks and open-ended queries.