Austria

We study continued training and supervised fine-tuning (SFT) of a language model (LM) to make effective use of long-context information. We first establish a reliable evaluation protocol to guide model development—instead of perplexity or simple needle-in-a-haystack (NIAH) tests, we use a broad set of long-context downstream tasks, and we evaluate models after SFT as this better reveals long-context abilities. Supported by our robust evaluations, we run thorough experiments to decide the data mix for continued pre-training, the instruction tuning dataset, and many other design choices such as position extrapolation. We find that (1) code repositories and books are excellent sources of long data, but it is crucial to combine them with high-quality short-context data; (2) training with a sequence length beyond the evaluation length boosts long-context performance; (3) for SFT, using only short instruction datasets yields strong performance on long-context tasks. Our final model, ProLong-8B, which is initialized from Llama-3 and trained on 40B tokens, demonstrates state-of-the-art long-context performance among similarly sized models at a length of 128K. ProLong outperforms Llama-3.1-8B-Instruct on the majority of long-context tasks despite using only 5% as many tokens during long-context training. Additionally, ProLong can effectively process up to 512K tokens, one of the longest context windows of publicly available LMs.

ACL 2025

How to Train Long-Context Language Models (Effectively)

data engineering

long-context

llms

continual training

## 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.

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)

Large language models (LLMs) have demonstrated remarkable capabilities across various domains, yet their application to relational deep learning (RDL) remains underexplored. Existing approaches adapt LLMs by traversing relational links between entities in a database and converting the structured data into flat text documents, but this text-based serialization disregards critical relational structures, introduces redundancy, and often exceeds standard LLM context lengths. We introduce Rel-LLM, a novel architecture that employs a graph neural network (GNN) based encoder to create structured relational prompts for LLMs within a retrieval-augmented generation (RAG) framework. Unlike traditional text-based serialization approaches, our method preserves the inherent relational structure of databases while enabling LLMs to effectively process and reason over complex entity relationships. 
Specifically, the GNN encoder extracts a local subgraph around an entity to build feature representations that contain relevant entity relationships and temporal dependencies. These representations are transformed into structured prompts using a denormalization process, effectively allowing the LLM to reason over relational structures. Through extensive experiments, we demonstrate that Rel-LLM outperforms existing methods on key RDL tasks, offering a scalable and efficient approach to integrating LLMs with structured data sources. Code is available at~\url{https://github.com/smiles724/Rel-LLM}.

Large Language Models are Good Relational Learners

Recent findings reveal that much of the knowledge in a Transformer-based Large Language Model (LLM) is encoded in its feed-forward (FFN) layers, where each FNN layer can be interpreted as the summation of sub-updates, each corresponding to a weighted column vector from the FFN's value parameter matrix that often encodes human-interpretable concepts. In light of this, we hypothesize that model performance and behaviors can be further enhanced and controlled by modulating the contributions of these sub-updates based on their relevance to the input or target output style, and propose LLMBraces, a novel and efficient method that computes relevance scores associated with value vectors in FFN layers and leverages these scores to dynamically adjust the contribution of sub-updates. By optimizing sub-update contributions, LLMBraces refines the prediction process, leading to more accurate and reliable outputs, much like a `brace' providing support and stability. Moreover, LLMBraces can be extended to support conditional control over generation characteristics, such as sentiment, thereby offering fine-grained steering of LLM outputs. Extensive experiments on various LLMs—including Qwen2.5-1.5B, Llama2-7B, and Llama3-8B—demonstrate that LLMBraces outperforms baseline approaches in both fine-tuning and zero-shot settings while requiring significantly fewer tunable parameters, up to 75% fewer compared to LoRA. Furthermore, LLMBraces excels in sentiment-controlled generation and toxicity reduction, highlighting its potential for flexible, controlled text generation across applications.

LLM Braces: Straightening Out LLM Predictions with Relevant Sub-Updates

Entity linking involves normalizing a mention in medical text to a unique identifier in a knowledge base, such as UMLS or MeSH. Most entity linkers follow a two-stage process: first, a candidate generation step selects high-quality candidates, and then a named entity disambiguation phase determines the best candidate for final linking. This study demonstrates that leveraging a large language model (LLM) as an entity disambiguator significantly enhances entity linking models' accuracy and recall. Specifically, the LLM disambiguator achieves remarkable improvements when applied to alias-matching entity linking methods. Without any fine-tuning, our approach establishes a new state-of-the-art (SOTA), surpassing previous methods on multiple prevalent biomedical datasets by up to 16 points in accuracy. We released our code on GitHub at https://github.com/ChristopheYe/llm_disambiguator

LLM as Entity Disambiguator for Biomedical Entity-Linking

Large language models (LLMs) play a crucial role in modern applications but face vulnerabilities related to the extraction of sensitive information. This includes unauthorized accesses to internal prompts and retrieval of personally identifiable information (PII) (e.g., in Retrieval-Augmented Generation based agentic applications). We examine these vulnerabilities in a question-answering (QA) setting where LLMs use retrieved documents or training knowledge as few-shot prompts. Although these documents remain confidential under normal use, adversaries can manipulate input queries to extract private content. In this paper, we propose a novel attack method by exploiting the model’s lower-ranked output tokens to leak sensitive information. We systematically evaluate our method, demonstrating its effectiveness in both the agentic application privacy extraction setting and the direct training data extraction. These findings reveal critical privacy risks in LLMs and emphasize the urgent need for enhanced safeguards against information leakage.

Exploiting the Shadows: Unveiling Privacy Leaks through Lower-Ranked Tokens in Large Language Models

Existing evaluation frameworks for retrieval-augmented generation (RAG) systems focus on answerable queries, but they overlook the importance of appropriately rejecting unanswerable requests. In this paper, we introduce UAEval4RAG, a comprehensive evaluation framework designed to evaluate whether RAG systems effectively handle unanswerable queries specific to a given knowledge base. We first define a taxonomy with six unanswerable categories, and UAEval4RAG automatically synthesizes diverse and challenging queries for any given knowledge base and evaluate the RAG systems with unanswered ratio and acceptable ratio metrics. We also conduct experiments with various RAG components and prompting strategies across four datasets, which reveals that due to varying knowledge distribution across datasets, no single configuration consistently delivers optimal performance on both answerable and unanswerable requests across different knowledge bases. Our findings highlight the critical role of component selection and prompt design in optimizing RAG systems to balance the accuracy of answerable queries with high rejection rates of unanswerable ones. UAEval4RAG provides valuable insights and tools for developing more robust and reliable RAG systems.

Unanswerability Evaluation for Retrieval Augmented Generation

We introduce SciVer, the first benchmark specifically designed to evaluate the ability of foundation models to verify claims within a multimodal scientific context.
SciVer consists of 3,000 expert-annotated examples over 1,113 scientific papers, covering four subsets, each representing a common reasoning type in multimodal scientific claim verification. 
To enable fine-grained evaluation, each example includes expert-annotated supporting evidence.
We assess the performance of 21 state-of-the-art multimodal foundation models, including o4-mini, Gemini-2.5-Flash, Llama-3.2-Vision, and Qwen2.5-VL. 
Our experiment reveals a substantial performance gap between these models and human experts on SciVer.
Through an in-depth analysis of retrieval-augmented generation (RAG), and human-conducted error evaluations, we identify critical limitations in current open-source models, offering key insights to advance models' comprehension and reasoning in multimodal scientific literature tasks.

SciVer: Evaluating Foundation Models for Multimodal Scientific Claim Verification

Large Language Models (LLMs) have shown remarkable capabilities as autonomous agents; yet existing benchmarks either focus on single-agent tasks or are confined to narrow domains, failing to capture the dynamics of multi-agent coordination and competition. In this paper, we introduce MultiAgentBench, a comprehensive benchmark designed to evaluate LLM-based multi-agent systems across diverse, interactive scenarios. Our framework measures not only task completion but also the quality of collaboration and competition using novel, milestone-based key performance indicators. Moreover, we evaluate various coordination protocols (including star, chain, tree, and graph topologies) and innovative strategies such as group discussion and cognitive planning. Notably, cognitive planning improves milestone achievement rates by 3%. Code and dataset will be made publicly available. Code and datasets are publicavailable at 
https://github.com/ulab-uiuc/MARBLE

MultiAgentBench : Evaluating the Collaboration and Competition of LLM agents

The proliferation of large language models (LLMs) has introduced unprecedented challenges in fake news detection due to benchmark data contamination (BDC), where evaluation benchmarks are inadvertently memorized during the pre-training, leading to the inflated performance metrics. Traditional evaluation paradigms, reliant on static datasets and closed-world assumptions, fail to account the BDC risk in large-scale pre-training of current LLMs. This paper introduces TripleFact, a novel evaluation framework for fake news detection task, which designed to mitigate BDC risk while prioritizing real-world applicability. TripleFact integrates three components: (1) Human-Adversarial Preference Testing (HAPT) to assess robustness against human-crafted misinformation, (2) Real-Time Web Agent with Asynchronous Validation (RTW-AV) to evaluate temporal generalization using dynamically sourced claims, and (3) Entity-Controlled Virtual Environment (ECVE) to eliminate entity-specific biases. Through experiments on 17 state-of-the-art LLMs, including GPT, LLaMA, and DeepSeek variants, TripleFact demonstrates superior contamination resistance compared to traditional benchmarks. Results reveal that BDC artificially inflates performance by up to 23\% in conventional evaluations, while TripleFact Score (TFS) remain stable within 4\% absolute error under controlled contamination. The framework's ability to disentangle genuine detection capabilities from memorization artifacts underscores its potential as a fake news detection benchmark for the LLM era.

TripleFact: Defending Data Contamination in the Evaluation of LLM-driven Fake News Detection

We present a hierarchy of natural language understanding abilities and argue for the importance of moving beyond assessments of understanding at the lexical and sentence levels to the discourse level. We propose the task of anaphora accessibility as a diagnostic for assessing discourse understanding, and to this end, present an evaluation dataset inspired by theoretical research in dynamic semantics. We evaluate human and LLM performance on our dataset and find that LLMs and humans align on some tasks and diverge on others. Such divergence can be explained by LLMs' reliance on specific lexical items during language comprehension, in contrast to human sensitivity to structural abstractions.

Meaning Beyond Truth Conditions: Evaluating Discourse Level Understanding via Anaphora Accessibility

The advent of 1-bit large language models (LLMs), led by BitNet b1.58, has spurred interest in ternary LLMs. Despite this, research and practical applications focusing on efficient edge inference for ternary LLMs remain scarce. To bridge this gap, we introduce Bitnet.cpp, an inference system optimized for BitNet b1.58 and ternary LLMs. Given that mixed-precision matrix multiplication (mpGEMM) constitutes the bulk of inference time in ternary LLMs, Bitnet.cpp incorporates a novel mpGEMM library to facilitate sub-2-bits-per-weight, efficient and lossless inference. The library features two core solutions: Ternary Lookup Table (TL), which addresses spatial inefficiencies of previous bit-wise methods, and Int2 with a Scale (I2_S), which ensures lossless edge inference, both enabling high-speed inference. Our experiments show that Bitnet.cpp achieves up to a 6.25x increase in speed over full-precision baselines and up to 2.32x over low-bit baselines, setting new benchmarks in the field. Additionally, we expand TL to element-wise lookup table (ELUT) for low-bit LLMs in the appendix, presenting both theoretical and empirical evidence of its considerable potential. Bitnet.cpp is publicly available at https://github.com/microsoft/BitNet/tree/paper, offering a sophisticated solution for the efficient and practical deployment of edge LLMs.

Downloads

Next from ACL 2025

Large Language Models are Good Relational Learners

Similar lecture

When Long Helps Short: How Context Length in Supervised Fine-tuning Affects Behavior of Large Language Models