Austria

Direct Preference Optimization (DPO) has recently emerged as an efficient and effective method for aligning large language models with human preferences. However, constructing high-quality preference datasets remains challenging, often necessitating expensive manual or powerful LM annotations. Additionally, standard DPO exhibits suboptimal performance in complex reasoning tasks, such as mathematical and code reasoning. In this paper, we introduce an approach to collect preference pairs through iterative sampling and execution feedback, tailored to the current learning state (\emph{e.g.} well-learned, mis-learned, and unlearned) of the policy model. To alleviate the failures of DPO and improve its applicability in reasoning tasks, we propose {\ours}, an iterative uncertainty-aware preference optimization method that achieves fine-grained preference control by assessing model confidence. We validate our approach across three reasoning tasks, incorporating five established reasoning datasets and one self-curated dataset. Our experimental results demonstrate an overall improvement of 3.6% over the standard DPO method and show the model exhibits promising generalizability.

ACL 2025

Uncertainty-Aware Iterative Preference Optimization for Enhanced LLM Reasoning

mathematical nlp

code generation and understanding

alignment

uncertainty

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community." 

*Message from the General Chair, Dirk Hovy*

EMNLP 2025

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

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)

Large language models (LLMs) have shown great potential across various industries due to their remarkable ability to generalize through instruction tuning. However, the limited availability of domain-specific data significantly hampers their performance on specialized tasks. While existing methods primarily focus on selecting training data from general datasets that are similar to the target domain, they often fail to consider the joint distribution of instructions, resulting in inefficient learning and suboptimal knowledge transfer. To address these challenges, we introduce **G2IS** (**G**radient-based **G**raph **I**nstruction **S**election), a novel method that constructs a mixed gradient-based instruction graph to capture the joint distribution and interdependencies among instructions. By accounting for the relationships between instructions, G2IS improves domain adaptation efficiency. Additionally, we propose a gradient walk algorithm to refine the data selection process, enhancing both training effectiveness and efficiency. Our experiments demonstrate that G2IS outperforms traditional methods across various domain adaptation tasks, yielding significant performance gains, particularly in complex, data-scarce scenarios. These results underscore the potential of G2IS in advancing the development of large, domain-specific models.

Beyond Similarity: A Gradient-based Graph Method for Instruction Tuning Data Selection

Vision-language retrieval-augmented generation (RAG) has become an effective approach for tackling Knowledge-Based Visual Question Answering (KB-VQA), which requires external knowledge beyond the visual content presented in images. The effectiveness of Vision-language RAG systems hinges on multimodal retrieval, which is inherently challenging due to the diverse modalities and knowledge granularities in both queries and knowledge bases. Existing methods have not fully tapped into the potential interplay between these elements. We propose a multimodal RAG system featuring a coarse-to-fine, multi-step retrieval that harmonizes multiple granularities and modalities to enhance efficacy. Our system begins with a broad initial search aligning knowledge granularity for cross-modal retrieval, followed by a multimodal fusion reranking to capture the nuanced multimodal information for top entity selection. A text reranker then filters out the most relevant fine-grained section for augmented generation. Extensive experiments on the InfoSeek and Encyclopedic-VQA benchmarks show our method achieves state-of-the-art retrieval performance and highly competitive answering results, underscoring its effectiveness in advancing KB-VQA systems. Our code can be found at https://github.com/ChaoLinAViy/OMGM.

OMGM: Orchestrate Multiple Granularities and Modalities for Efficient Multimodal Retrieval

In the era of large models, content generation is gradually shifting to Personalized Generation (PGen), tailoring content to individual preferences and needs. This paper presents the first comprehensive survey on PGen, investigating existing research in this rapidly growing field. We conceptualize PGen from a unified perspective, systematically formalizing its key components, core objectives, and abstract workflows. Based on this unified perspective, we propose a multi-level taxonomy, offering an in-depth review of technical advancements, commonly used datasets, and evaluation metrics across multiple modalities, personalized contexts, and tasks. Moreover, we envision the potential applications of PGen and highlight open challenges and promising directions for future exploration. By bridging PGen research across multiple modalities, this survey serves as a valuable resource for fostering knowledge sharing and interdisciplinary collaboration, ultimately contributing to a more personalized digital landscape.

Personalized Generation In Large Model Era: A Survey

Multimodal Large Language Models (MLLMs) have serious security vulnerabilities. While safety alignment using multimodal datasets consisting of text and data of additional modalities can effectively enhance MLLM's security, it is costly to construct these datasets. Existing low-resource security alignment methods, including textual alignment, have been found to struggle with the security risks posed by additional modalities. To address this, we propose Synthetic Embedding augmented safety Alignment (SEA), which optimizes embeddings of additional modality through gradient updates to expand textual datasets. This enables multimodal safety alignment training even when only textual data is available. Extensive experiments on image, video, and audio-based MLLMs demonstrate that SEA can synthesize a high-quality embedding on a single RTX3090 GPU within 24 seconds. SEA significantly improves the security of MLLMs when faced with threats from additional modalities. To assess the security risks introduced by video and audio, we also introduced a new benchmark called VA-SafetyBench. High attack success rates across multiple MLLMs validate its challenge. Our code and data will be available at https://github.com/ZeroNLP/SEA.

SEA: Low-Resource Safety Alignment for Multimodal Large Language Models via Synthetic Embeddings

Prompt compression is important for large language models (LLMs) to increase inference speed, reduce costs, and improve user experience. However, current methods face challenges such as low compression ratios and potential training-test overlap during evaluation. To address these issues, we propose 500xCompressor, a method that compresses natural language contexts into a minimum of one special token and demonstrates strong generalization ability. The 500xCompressor introduces approximately 0.3% additional parameters and achieves compression ratios ranging from 6x to 500x, achieving 27-90% reduction in calculations and 55-83% memory savings when generating 100-400 tokens for new and reused prompts at 500x compression, while retaining 70-74% (F1) and 77-84% (Exact Match) of the LLM capabilities compared to using non-compressed prompts. It is designed to compress any text, answer various types of questions, and can be utilized by the original LLM without requiring fine-tuning. Initially, 500xCompressor was pretrained on the ArxivCorpus, followed by fine-tuning on the ArxivQA dataset, and subsequently evaluated on strictly unseen and cross-domain question answering (QA) datasets. This study shows that KV values outperform embeddings in preserving information at high compression ratios. The highly compressive nature of natural language prompts, even for detailed information, suggests potential for future applications and the development of a new LLM language.

500xCompressor: Generalized Prompt Compression for Large Language Models

Language models (LMs) rely on their parametric knowledge augmented with relevant contextual knowledge for certain tasks, such as question answering. However, the contextual knowledge can contain private information that may be leaked when answering queries, and estimating this privacy leakage is not well understood. A straightforward approach of directly comparing an LM's output to the contexts can overestimate the privacy risk, since the LM's parametric knowledge might already contain the augmented contextual knowledge. To this end, we introduce \emph{context influence}, a metric that builds on differential privacy, a widely-adopted privacy notion, to estimate the privacy leakage of contextual knowledge during decoding. Our approach effectively measures how each subset of the context influences an LM's response while separating the specific parametric knowledge of the LM. Using our context influence metric, we demonstrate that context privacy leakage occurs when contextual knowledge is out of distribution with respect to parametric knowledge. Moreover, we experimentally demonstrate how context influence properly attributes the privacy leakage to augmented contexts, and we evaluate how factors-- such as model size, context size, generation position, etc.-- affect context privacy leakage. The practical implications of our results will inform practitioners of the privacy risk associated with augmented contextual knowledge.

Estimating Privacy Leakage of Augmented Contextual Knowledge in Language Models

Transformer models have achieved remarkable performance in many formal reasoning tasks. Nonetheless, the extent of their comprehension pertaining to logical semantics and rules of inference remains somewhat uncertain. Evaluating such understanding necessitates a rigorous examination of these models' generalisation capacity to out-of-distribution data. In this study, we probe the generalisation prowess of Transformer models with respect to the hitherto unexplored domain of numerical satisfiability problems. Our investigation reveals that Transformers exhibit minimal scale and noise invariance, alongside limited vocabulary and number invariance. However, even when Transformer models experience a notable decline in performance on out-of-distribution test sets, they often still surpass the random baseline by a considerable margin.

Unravelling the Logic: Investigating the Generalisation of Transformers in Numerical Satisfiability Problems

Quality estimation (QE)---the automatic assessment of translation quality---has recently become crucial across several stages of the translation pipeline, from data curation to training and decoding. While QE metrics have been optimized to align with human judgments, whether they encode social biases has been largely overlooked. Biased QE risks favoring certain demographic groups over others, e.g., by exacerbating gaps in visibility and usability. This paper defines and investigates gender bias of QE metrics and discusses its downstream implications for machine translation (MT). Experiments with state-of-the-art QE metrics across multiple domains, datasets, and languages reveal significant bias. When a human entity's gender in the source is undisclosed, masculine-inflected translations score higher than feminine-inflected ones, and gender-neutral translations are penalized. Even when contextual cues disambiguate gender, using context-aware QE metrics leads to more errors in selecting the correct translation inflection for feminine referents than for masculine ones. Moreover, a biased QE metric affects data filtering and quality-aware decoding. Our findings underscore the need for a renewed focus on developing and evaluating QE metrics centered on gender.

Watching the Watchers: Exposing Gender Disparities in Machine Translation Quality Estimation

Inference-Time Scaling has been critical to the success of recent models such as OpenAI o1 and DeepSeek R1. However, many techniques used to train models for inference-time scaling require tasks to have answers that can be verified, limiting their application to domains such as math, coding and logical reasoning. We take inspiration from how humans make first attempts, ask for detailed feedback from others and make improvements based on such feedback across a wide spectrum of open-ended endeavors. To this end, we collect HelpSteer3 data to train dedicated Feedback and Edit Models that are capable of performing inference-time scaling for open-ended general-domain tasks. In our setup, one model generates an initial response, which are given feedback by a second model, that are then used by a third model to edit the response. We show that performance on Arena Hard, a benchmark strongly predictive of Chatbot Arena Elo can be boosted by scaling the number of initial response drafts, effective feedback and edited responses. When scaled optimally, our setup based on 70B models from the Llama 3 family can reach SoTA performance on Arena Hard at 92.7 as of 5 Mar 2025, surpassing OpenAI o1-preview-2024-09-12 with 90.4 and DeepSeek R1 with 92.3.

Dedicated Feedback and Edit Models Empower Inference-Time Scaling for Open-Ended General-Domain Tasks

Automating scientific research is considered the final frontier of science. Recently, several papers claim autonomous research agents can generate novel research ideas. Amidst the prevailing optimism, we document a critical concern: a considerable fraction of such research documents are smartly plagiarized. Unlike past efforts where experts evaluate the novelty and feasibility of research ideas, we request $13$ experts to operate under a different situational logic: to identify similarities between LLM-generated research documents and existing work. Concerningly, the experts identify $24$% of the $50$ evaluated research documents to be either paraphrased (with one-to-one methodological mapping), or significantly borrowed from existing work. These reported instances are cross-verified by authors of the source papers. Experts find an additional $32$% ideas to partially overlap with prior work, and a small fraction to be completely original. Problematically, these LLM-generated research documents do not acknowledge original sources, and bypass inbuilt plagiarism detectors. Lastly, through controlled experiments we show that automated plagiarism detectors are inadequate at catching plagiarized ideas from such systems. We recommend a careful assessment of LLM-generated research, and discuss the implications of our findings on academic publishing.

Downloads

Next from ACL 2025

Beyond Similarity: A Gradient-based Graph Method for Instruction Tuning Data Selection

Similar lecture

Expectation Preference Optimization: Reliable Preference Estimation for Improving the Reasoning Capability of Large Language Models