Austria

Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes often relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers leads to poor generalization on reasoning tasks that require more detailed explanations. To address this limitation, we propose a two-stage post-training strategy that extends the usage of short answer data for enhanced CoT reasoning. First, we augment short answers with CoT reasoning generated by GPT-4o, enhancing the VLM’s CoT capabilities through fine-tuning. Second, we leverage short answers as outcome rewards for reinforcement learning. Specifically, short answers are used as correctness indicators to construct positive (correct) and negative (incorrect) pairs from model-generated reasoning chains. These pairs are then used to calibrate the model’s reasoning via Direct Preference Optimization. Our experiments show significant improvements in CoT reasoning on benchmark datasets, along with enhanced generalization to direct answer prediction. This work provides a critical data resource for VLM CoT training and demonstrates the effectiveness of outcome rewards for multimodal models post-training.

ACL 2025

Improve Vision Language Model Chain-of-thought Reasoning

vision lanugage model

chain-of-thought reasoning

reinforcement learning

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

ACL 2024

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

technical paper

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2025". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you need to log in with the **email address you registered with**. 

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Non-collaborative dialogue involves two participants with conflicting interests engaging in a multi-round dialogue to achieve their own goals. Strategy planning is the key to guiding both participants towards a consensus. Most LLMs-based methods use stimulus prompts or external strategy planners for strategy planning. However, stimulus prompts fail to teach LLMs to plan dialogue strategies explicitly. Moreover, training external strategy planners doesn't fully account for adversarial interactions, thereby limiting their effectiveness against tough resisters. In this paper, to mitigate the above issues, we propose $\textbf{GAIA}$, a $\textbf{G}$ame-based $\textbf{A}$dversarial self-play $\textbf{I}$nter$\textbf{A}$ctive training paradigm, which constructs an adversarial two-player (a persuader and a resister) zero-sum game and guides the game to approximate Nash Equilibrium (NE) via reinforcement learning (RL) for the non-collaborative dialogues. First, we design a Chain-of-Mind prompt to reason the resister’s dialogue act step-by-step to plan the persuasive strategies. Secondly, to adversarially improve the persuader, we construct diverse resistant planners and theoretically improve the persuader's optimal lower bound. Finally, we iteratively optimise their policies via adversarial self-play interactive RL and design an $\epsilon$-NE verification algorithm to approximate the game's NE. Experiments on three datasets show that our model obtains state-of-the-art performance.

Battling against Tough Resister: Strategy Planning with Adversarial Game for Non-collaborative Dialogues

Large Vision-Language Models (LVLMs) have shown exceptional performance in multimodal tasks, but their effectiveness in complex visual reasoning is still constrained, especially when employing Chain-of-Thought prompting techniques. In this paper, we propose VReST, a novel training-free approach that enhances Reasoning in LVLMs through Monte Carlo Tree Search and Self-Reward mechanisms. VReST meticulously traverses the reasoning landscape by establishing a search tree, where each node encapsulates a reasoning step, and each path delineates a comprehensive reasoning sequence. Our innovative multimodal Self-Reward mechanism assesses the quality of reasoning steps by integrating the utility of sub-questions, answer correctness, and the relevance of vision-language clues, all without the need for additional models. VReST surpasses current prompting methods and secures state-of-the-art performance across three multimodal mathematical reasoning benchmarks. Furthermore, it substantiates the efficacy of test-time scaling laws in multimodal tasks, offering a promising direction for future research.

VReST: Enhancing Reasoning in Large Vision-Language Models through Tree Search and Self-Reward Mechanism

Multi-hop question answering (QA) involves finding multiple relevant passages and performing step-by-step reasoning to answer complex questions. Previous works on multi-hop QA employ specific methods from different modeling perspectives based on large language models (LLMs), regardless of the question types. In this paper, we first conduct an in-depth analysis of public multi-hop QA benchmarks, dividing the questions into four types and evaluating five types of cutting-edge methods for multi-hop QA: Chain-of-Thought (CoT), Single-step, Iterative-step, Sub-step, and Adaptive-step. We find that different types of multi-hop questions have varying degrees of sensitivity to different types of methods. Thus, we propose a Bi-levEL muLti-agEnt reasoning (BELLE) framework to address multi-hop QA by specifically focusing on the correspondence between question types and methods, where each type of method is regarded as an ''operator'' by prompting LLMs differently. The first level of BELLE includes multiple agents that debate to obtain an executive plan of combined ''operators'' to address the multi-hop QA task comprehensively. During the debate, in addition to the basic roles of affirmative debater, negative debater, and judge, at the second level, we further leverage fast and slow debaters to monitor whether changes in viewpoints are reasonable. Extensive experiments demonstrate that BELLE significantly outperforms strong baselines in various datasets. Additionally, the model consumption of BELLE is higher cost-effectiveness than that of single models in more complex multi-hop QA scenarios.

BELLE: A Bi-Level Multi-Agent Reasoning Framework for Multi-Hop Question Answering

Fine-grained emotion classification (FEC) aims to analyze speakers' utterances and distinguish dozens of emotions with subtle differences, allowing for a more nuanced understanding of human emotional states. However, compared to traditional coarse-grained emotion classification, two difficulties arise as the granularity of emotions becomes finer, i.e., the presence of closely confusable emotions which are hard to distinguish, and the biased performance caused by long-tailed emotions. Although addressing both difficulties is vital to FEC, previous studies have predominantly focused on dealing with only one of them. In this paper, we propose TACO, a novel triple-view framework that treats FEC as an instance-label (i.e., utterance-emotion) joint embedding learning problem to tackle both difficulties concurrently by considering three complementary views. Specifically, we design a clustering-guided contrastive loss, which incorporates clustering techniques to guide the contrastive learning process and facilitate more discriminative instance embeddings. Additionally, we introduce the emotion label description as a helpful resource to refine label embeddings and mitigate the poor performance towards under-represented (i.e., long-tailed) emotions. Extensive experiments on two widely-used benchmark datasets demonstrate that our proposed TACO achieves substantial and consistent improvements compared to other competitive baseline methods.

A Triple-View Framework for Fine-Grained Emotion Classification with Clustering-Guided Contrastive Learning

When exposed to complex queries containing multiple conditions, today’s large language models (LLMs) tend to produce responses that only partially satisfy the query while neglecting certain conditions. We, therefore, introduce the concept of Intent Hallucination, a phenomenon where LLMs either omit (failing to address certain parts) or misinterpret (responding to invented query parts) elements of the given query, leading to responses misaligned with the original query. To systematically evaluate intent hallucination, we introduce FAITHQA, a novel benchmark for intent hallucination that contains 20,068 problems, covering both query-only and retrieval-augmented generation (RAG) setups with varying topics and difficulty. FAITHQA is the first hallucination benchmark that goes beyond factual verification, tailored to identify the fundamental cause of intent hallucination. By evaluating various LLMs on FAITHQA, we find that (1) intent hallucination is a common issue even for state-of-the-art models, and (2) such a phenomenon stems from omission or misinterpretation of LLMs. To facilitate future research, we introduce an automatic LLM generation evaluation metric, named INTENT CONSTRAINT, for detecting intent hallucination. Human evaluation results demonstrate that INTENT CONSTRAINT is closer to human performance for intent hallucination compared to baselines.

Beyond Facts: Evaluating Intent Hallucination in Large Language Models

Ensuring the reliability of Large Language Models (LLMs) in complex reasoning tasks remains a formidable challenge, particularly in scenarios that demand precise mathematical calculations and knowledge-intensive open-domain generation. In this work, we introduce an uncertainty-aware framework designed to enhance the accuracy of LLM responses by systematically incorporating model confidence at critical decision points. We propose an approach that encourages multi-step reasoning in LLMs and quantify the confidence of intermediate answers such as numerical results in mathematical reasoning and proper nouns in open-domain generation. Then, the overall confidence of each reasoning chain is evaluated based on confidence of these critical intermediate steps. Finally, we aggregate the answer of generated response paths in a way that reflects the reliability of each generated content (as opposed to self-consistency in which each generated chain contributes equally to majority voting). We conducted extensive experiments in five datasets, three mathematical datasets and two open-domain datasets, using four LLMs. The results consistently validate the effectiveness of our novel confidence-aggregation method, leading to an accuracy improvement of up to 7.4\% and 5.8\% over baseline approaches in math and open-domain generation tasks, respectively. Code is publicly available at https://github.com/sharif-ml-lab/CER.

CER: Confidence Enhanced Reasoning in LLMs

With the increasing prominence of large language models (LLMs), evaluating their text-generation capabilities has become an essential research challenge. Although LLM-based evaluation methods exhibit robust performance, the inherent stochastic nature of the LLM generation process introduces a degree of uncertainty in alignment with human preferences. To address this limitation, we propose Semantic-Eval, the first training-free framework designed to assess LLM-generated text based on semantic understanding. This framework computes semantic similarity between pairwise texts to evaluate the interdependence of semantic units, integrating a graph-based weighting mechanism to account for the differential contributions of individual sentences. A pre-trained natural language inference (NLI) model is also incorporated to mitigate potential semantic relationship biases. We evaluate Semantic-Eval across eight datasets that encompass four common NLP tasks. The experimental results indicate that Semantic-Eval surpasses traditional N-gram and BERT-based evaluation metrics, aligning more closely with human judgments and demonstrating a higher correlation than smaller LLMs. However, it slightly lags behind GPT-4. Finally, we demonstrate the effectiveness of Semantic-Eval in evaluating the generation quality of 13 large language models. The code is publicly available at https://github.com/LssTry/Semantic-Eval.

Semantic-Eval : A Semantic Comprehension Evaluation Framework for Large Language Models Generation without Training

As speech generation technology advances, the risk of misuse through deepfake audio has become a pressing concern, which underscores the critical need for robust detection systems. However, many existing speech deepfake datasets are limited in scale and diversity, making it challenging to train models that can generalize well to unseen deepfakes. To address these gaps, we introduce SpeechFake, a large-scale dataset designed specifically for speech deepfake detection. SpeechFake includes over 3 million deepfake samples, totaling more than 3,000 hours of audio, generated using 40 different speech synthesis tools. The dataset encompasses a wide range of generation techniques, including text-to-speech, voice conversion, and neural vocoder, incorporating the latest cutting-edge methods. It also provides multilingual support, spanning 46 languages. In this paper, we offer a detailed overview of the dataset’s creation, composition, and statistics. We also present baseline results by training detection models on SpeechFake, demonstrating strong performance on both its own test sets and various unseen test sets. Additionally, we conduct experiments to rigorously explore how generation methods, language diversity, and speaker variation affect detection performance. We believe SpeechFake will be a valuable resource for advancing speech deepfake detection and developing more robust models for evolving generation techniques.

SpeechFake: A Large-Scale Multilingual Speech Deepfake Dataset Incorporating Cutting-Edge Generation Methods

Neural machine translation (NMT) has advanced significantly, yet challenges remain in adapting to new domains . In scenarios where bilingual data is limited, this issue is further exacerbated. To address this, we propose $k$NN-LM-NMT, a method that leverages semantically similar target language sentences in the $k$NN framework. Our approach generates a probability distribution over these sentences during decoding, and this distribution is then interpolated with the NMT model's distribution. Additionally, we introduce an $n$-gram-based approach to focus on similar fragments, enabling the model to avoid the noise introduced by the non-similar parts. To enhance accuracy, we further incorporate cross-lingual retrieval similarity to refine the $k$NN probability distribution. Extensive experiments on multi-domain datasets demonstrate significant performance improvements in both high-resource and low-resource scenarios. Our approach effectively extracts translation knowledge from limited target domain data, and well benefits from large-scale monolingual data for robust context representation.

Enhancing Neural Machine Translation Through Target Language Data: A kNN-LM Approach for Domain Adaptation

In AI-facilitated teaching, leveraging various query styles to interpret abstract text descriptions is crucial for ensuring high-quality teaching. However, current retrieval models primarily focus on natural text-image retrieval, making them insufficiently tailored to educational scenarios due to the ambiguities in the retrieval process. In this paper, we propose a diverse expression retrieval task tailored to educational scenarios, supporting retrieval based on multiple query styles and expressions. We introduce the STEM Education Retrieval Dataset (SER), which contains over 24,000 query pairs of different styles, and the Uni-Retrieval, an efficient and style-diversified retrieval vision-language model based on prompt tuning. Uni-Retrieval extracts query style features as prototypes and builds a continuously updated Prompt Bank containing prompt tokens for diverse queries. This bank can updated during test time to represent domain-specific knowledge for different subject retrieval scenarios. Our framework demonstrates scalability and robustness by dynamically retrieving prompt tokens based on prototype similarity, effectively facilitating learning for unknown queries. Experimental results indicate that Uni-Retrieval outperforms existing retrieval models in most retrieval tasks.

Downloads

Next from ACL 2025

Battling against Tough Resister: Strategy Planning with Adversarial Game for Non-collaborative Dialogues

Similar lecture

Direct Evaluation of Chain-of-Thought in Multi-hop Reasoning with Knowledge Graphs