Austria

Knowledge distillation (KD) is a widely used approach for BERT compression, where a larger BERT model serves as a teacher to transfer knowledge to a smaller student model. Prior works have found that distilling a larger BERT with superior performance may degrade student&#39;s performance than a smaller BERT. In this paper, we investigate the limitations of existing KD methods for larger BERT models. Through Canonical Correlation Analysis, we identify that these methods fail to fully exploit the potential advantages of larger teachers. To address this, we propose an improved distillation approach that effectively enhances knowledge transfer. Comprehensive experiments demonstrate the effectiveness of our method in enabling larger BERT models to distill knowledge more efficiently.

ACL 2025

Maximizing the Effectiveness of Larger BERT Models for Compression

nlp in resource-constrained settings

data-efficient training

bert compression

knowledge distillation

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

AAAI 2021

Knowledge distillation (KD) is a widely used approach for BERT compression, where a larger BERT model serves as a teacher to transfer knowledge to a smaller student model. Prior works have found that distilling a larger BERT with superior performance may degrade student's performance than a smaller BERT. In this paper, we investigate the limitations of existing KD methods for larger BERT models. Through Canonical Correlation Analysis, we identify that these methods fail to fully exploit the potential advantages of larger teachers. To address this, we propose an improved distillation approach that effectively enhances knowledge transfer. Comprehensive experiments demonstrate the effectiveness of our method in enabling larger BERT models to distill knowledge more efficiently.

poster

### 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)

Recent advances in synthetic speech generation technology have facilitated the generation of high-quality synthetic (fake) speech that emulates human voices. These technologies pose a threat of misuse for identity theft and the spread of misinformation. Consequently, the misuse of such powerful technologies necessitates the development of robust and generalizable audio deepfake detection (ADD) and anti-spoofing models. However, such models are often linguistically biased. Consequently, the models trained on datasets in one language exhibit a low accuracy when evaluated on out-of-domain languages. Such biases reduce the usability of these models and highlight the urgent need for multilingual synthetic speech datasets for bias mitigation research. However, most available datasets are in English or Chinese. The dearth of multilingual synthetic datasets hinders multilingual ADD and anti-spoofing research. Furthermore, the problem intensifies in countries with rich linguistic diversity, such as India. Therefore, we introduce IndicSynth, which contains 4,000 hours of synthetic speech from 989 target speakers, including 456 females and 533 males for 12 low-resourced Indian languages. The dataset includes rich metadata covering gender details and target speaker identifiers. Experimental results demonstrate that IndicSynth is a valuable contribution to multilingual ADD and anti-spoofing research. The dataset can be accessed from https://github.com/vdivyas/IndicSynth.

IndicSynth: A Large-Scale Multilingual Synthetic Speech Dataset for Low-Resource Indian Languages

Multimodal retrieval has emerged as a promising yet challenging research direction in recent years. Most existing studies in multimodal retrieval focus on capturing information in multimodal data that is similar to their paired texts, but often ignores the complementary information contained in multimodal data. In this study, we propose CIEA, a novel multimodal retrieval approach that employs Complementary Information Extraction and Alignment, which transforms both text and images in documents into a unified latent space and features a complementary information extractor designed to identify and preserve differences in the image representations. We optimize CIEA using two complementary contrastive losses to ensure semantic integrity and effectively capture the complementary information contained in images. Extensive experiments demonstrate the effectiveness of CIEA, which achieves significant improvements over both divide-and-conquer models and universal dense retrieval models. We provide an ablation study, further discussions, and case studies to highlight the advancements achieved by CIEA. To promote further research in the community, we have released the source code at https://github.com/zengdlong/CIEA.

Enhancing Multimodal Retrieval via Complementary Information Extraction and Alignment

Although value-aligned language models (LMs) appear unbiased in explicit bias evaluations, they often exhibit stereotypes in implicit word association tasks, raising concerns about their fair usage. We investigate the mechanisms behind this discrepancy and find that alignment surprisingly amplifies implicit bias in model outputs. Specifically, we show that aligned LMs, unlike their unaligned counterparts, overlook racial concepts in early internal representations when the context is ambiguous. Not representing race likely fails to activate safety guardrails, leading to unintended biases. Inspired by this insight, we propose a new bias mitigation strategy that works by incentivizing the representation of racial concepts in the early model layers. In contrast to conventional mitigation methods of machine unlearning, our interventions find that steering the model to be more aware of racial concepts effectively mitigates implicit bias. Similar to race blindness in humans, ignoring racial nuances can inadvertently perpetuate subtle biases in LMs.

Aligned but Blind: Alignment Increases Implicit Bias by Reducing Awareness of Race

Knowledge distillation typically involves transferring knowledge from a Large Language Model (LLM) to a Smaller Language Model (SLM). However, in tasks like text matching, smaller fine-tuned models often produce more effective domain-specific representations as they focus on optimizing the similarity between input pairs. To combine the specialized strengths of small models with the rich semantic understanding of LLMs, we propose a flipped knowledge distillation paradigm, where the LLM learns from the SLM. To bridge the architectural gap between commonly used decoder-only LLMs and the encoder-based frameworks of smaller models, we reinterpret LLMs as encoder-decoder models using LoRA. In this setup, the encoder generates compressed text representations, while the decoder transforms them into the output space. During training, the encoder produces text representations and computes their similarities, which are then aligned with the similarity scores produced by the teacher model. We achieve this alignment using our proposed Margin-aware Contrastive Learning (MCL) approach. MCL ensures accurate similarity for both positive and negative pairs, while also adaptively handling differences within positive and negative samples. We validate the effectiveness of our approach on financial and healthcare benchmarks as well as real-world online applications. Our model has been fully deployed in an online application environment, demonstrating its practical utility.

Flipping Knowledge Distillation: Leveraging Small Models' Expertise to Enhance LLMs in Text Matching

This paper investigates the faithfulness of multimodal large language model (MLLM) agents in a graphical user interface (GUI) environment, aiming to address the research question of whether multimodal GUI agents can be distracted by environmental context. A general scenario is proposed where both the user and the agent are benign, and the environment, while not malicious, contains unrelated content. A wide range of MLLMs are evaluated as GUI agents using a simulated dataset, following three working patterns with different levels of perception. Experimental results reveal that even the most powerful models, whether generalist agents or specialist GUI agents, are susceptible to distractions. While recent studies predominantly focus on the helpfulness of agents, our findings first indicate that these agents are prone to environmental distractions. Furthermore, we implement an adversarial environment injection and analyze the approach to improve faithfulness, calling for a collective focus on this important topic.

Caution for the Environment: LLM Agents are Susceptible to Environmental Distractions

In this paper, we identify a critical problem, "lost-in-retrieval", in retrieval-augmented multi-hop question answering (QA): the key entities are missed in LLMs' sub-question decomposition. "Lost-in-retrieval" significantly degrades the retrieval performance, which disrupts the reasoning chain and leads to the incorrect answers. To resolve this problem, we propose a progressive retrieval and rewriting method, namely ChainRAG, which sequentially handles each sub-question by completing missing key entities and retrieving relevant sentences from a sentence graph for answer generation. Each step in our retrieval and rewriting process builds upon the previous one, creating a seamless chain that leads to accurate retrieval and answers. Finally, all retrieved sentences and sub-question answers are integrated to generate a comprehensive answer to the original question. We evaluate ChainRAG on three multi-hop QA datasets—MuSiQue, 2Wiki, and HotpotQA—using three large language models: GPT4o-mini, Qwen2.5-72B, and GLM-4-Plus. Empirical results demonstrate that ChainRAG consistently outperforms baselines in both effectiveness and efficiency.

Mitigating Lost-in-Retrieval Problems in Retrieval Augmented Multi-Hop Question Answering

The quality of Supervised Fine-Tuning (SFT) data plays a critical role in enhancing the conversational capabilities of Large Language Models (LLMs). However, the availability of high-quality human-annotated SFT data has become a significant bottleneck for LLMs, necessitating a greater reliance on synthetic training data. In this work, we introduce Condor, a two-stage synthetic data generation framework that incorporates World Knowledge Trees and Self-Reflection Refinement to produce high-quality SFT data at scale. Our experimental results demonstrate that a base model fine-tuned on only 20K Condor-generated samples achieves superior performance compared to instruct model trained with RLHF. The additional refinement stage in Condor further enables iterative self-improvement for LLMs at various scales (up to 72B), validating the effectiveness of our approach. Furthermore, our investigation into the scaling of synthetic data in post-training reveals substantial unexplored potential for performance improvements, opening promising avenues for future research.

Condor: Enhance LLM Alignment with Knowledge-Driven Data Synthesis and Refinement

Multi-agent collaborative tasks exhibit exceptional capabilities in natural language applications and generation. By prompting agents to assign clear roles, it is possible to facilitate cooperation and achieve complementary capabilities among LLMs. A common strategy involves adopting a relatively general role assignment mechanism, such as introducing a ``judge'' or a ``summarizer''. However, these approaches lack task-specific role customization based on task characteristics. Another strategy involves decomposing the task based on domain knowledge and task characteristics, followed by assigning appropriate roles according to LLMs' respective strengths, such as programmers and testers. However, in some given tasks, obtaining domain knowledge related to task characteristics and getting the strengths of different LLMs is hard. To solve these problems, we propose a Multi-LLM Cooperation (MLC) framework with automatic role assignment capabilities. The core idea of the MLC is to initialize role assignments randomly and then allow the role embeddings to be learned jointly with the downstream task. To capture the state transitions of multiple LLMs during turn-based speaking, the role embedding is sequence-aware. At the same time, to avoid role convergence, the role differentiation module in MLC encourages behavioral differentiation between LLMs while ensuring the LLM team consistency, guiding different LLMs to develop complementary strengths from the optimization level. Our experiments on seven datasets demonstrate that MLC significantly enhances collaboration and expertise, which collaboratively addresses multi-agent tasks.

Advancing Collaborative Debates with Role Differentiation through Multi-Agent Reinforcement Learning

Critical text assessment is at the core of many expert activities, such as fact-checking, peer review, and essay grading. Yet, existing work treats critical text assessment as a black box problem, limiting interpretability and human-AI collaboration. To close this gap, we introduce Structured Reasoning in Critical Text Assessment (STRICTA), a novel specification framework to model text assessment as an explicit, step-wise reasoning process. STRICTA breaks down the assessment into a graph of interconnected reasoning steps drawing on causality theory (Pearl, 1995). This graph is populated based on expert interaction data and used to study the assessment process and facilitate human-AI collaboration. We formally define STRICTA and apply it in a study on biomedical paper assessment, resulting in a dataset of over 4000 reasoning steps from roughly 40 biomedical experts on more than 20 papers. We use this dataset to empirically study expert reasoning in critical text assessment, and investigate if LLMs are able to imitate and support experts within these workflows. The resulting tools and datasets pave the way for studying collaborative expert-AI reasoning in text assessment, in peer review and beyond.

STRICTA: Structured Reasoning in Critical Text Assessment for Peer Review and Beyond

To meet the demands of content moderation, online platforms have resorted to automated systems. Newer forms of real-time engagement ($\textit{e.g.}$, users commenting on live streams) on platforms like Twitch exert additional pressures on the latency expected of such moderation systems. Despite their prevalence, relatively little is known about the effectiveness of these systems. In this paper, we conduct an audit of Twitch's automated moderation tool ($\texttt{AutoMod}$) to investigate its effectiveness in flagging hateful content. For our audit, we create streaming accounts to act as siloed test beds, and interface with the live chat using Twitch's APIs to send over $107,000$ comments collated from $4$ datasets. We measure $\texttt{AutoMod}$'s accuracy in flagging blatantly hateful content containing misogyny, racism, ableism and homophobia. Our experiments reveal that a large fraction of hateful messages, up to $94$% on some datasets, $\text{\textit{bypass moderation}}$. Contextual addition of slurs to these messages results in $100$% removal, revealing $\texttt{AutoMod}$'s reliance on slurs as a hate signal. We also find that contrary to Twitch's community guidelines, $\texttt{AutoMod}$ blocks up to $89.5$% of benign examples that use sensitive words in pedagogical or empowering contexts. Overall, our audit points to large gaps in $\texttt{AutoMod}$'s capabilities and underscores the importance for such systems to understand context effectively.

Downloads

Next from ACL 2025

IndicSynth: A Large-Scale Multilingual Synthetic Speech Dataset for Low-Resource Indian Languages

Similar lecture

ROSITA: Refined BERT Compression with Integrated Techniques