China

Recently, autonomous agents built on large language models (LLMs) have experienced significant development and are being deployed in real-world applications. Through the usage of tools, these systems can perform actions in the real world. Given the agents&#39; practical applications and ability to execute consequential actions, such autonomous systems can cause more severe damage than a standalone LLM if compromised. While some existing research has explored harmful actions by LLM agents, our study approaches the vulnerability from a different perspective. We introduce a new type of attack that causes malfunctions by misleading the agent into executing repetitive or irrelevant actions. Our experiments reveal that these attacks can induce failure rates exceeding 80% in multiple scenarios. Through attacks on implemented and deployable agents in multi-agent scenarios, we accentuate the realistic risks associated with these vulnerabilities. To mitigate such attacks, we propose self-examination defense methods. Our findings indicate these attacks are more difficult to detect compared to previous overtly harmful attacks, highlighting the substantial risks associated with this vulnerability.

EMNLP 2025

Breaking Agents: Compromising Autonomous LLM Agents Through Malfunction Amplification

agents

attack

security

### 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.

Recently, autonomous agents built on large language models (LLMs) have experienced significant development and are being deployed in real-world applications. Through the usage of tools, these systems can perform actions in the real world. Given the agents' practical applications and ability to execute consequential actions, such autonomous systems can cause more severe damage than a standalone LLM if compromised. While some existing research has explored harmful actions by LLM agents, our study approaches the vulnerability from a different perspective. We introduce a new type of attack that causes malfunctions by misleading the agent into executing repetitive or irrelevant actions. Our experiments reveal that these attacks can induce failure rates exceeding 80% in multiple scenarios. Through attacks on implemented and deployable agents in multi-agent scenarios, we accentuate the realistic risks associated with these vulnerabilities. To mitigate such attacks, we propose self-examination defense methods. Our findings indicate these attacks are more difficult to detect compared to previous overtly harmful attacks, highlighting the substantial risks associated with this vulnerability.

poster

## 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*

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

The detection of audio deepfakes (ADD) has become increasingly important due to the rapid evolution of generative speech models. However, progress in this field remains uneven across languages, particularly for low-resource languages like Portuguese, which lack high-quality datasets. In this paper, we introduce BRSpeech-DF, the first publicly available ADD dataset for Portuguese, encompassing both Brazilian and European variants. The dataset contains over 459,000 utterances, including a smaller portion of real speech from 62 speakers and a large collection of synthetic samples generated using multiple zero-shot text-to-speech (TTS) models, each conditioned on the original speaker's voice. By providing this resource, our objective is to support the development of robust, multilingual detection systems, thereby advancing equity in speech forensics and security research. BRSpeech-DF addresses a significant gap in annotated data for underrepresented languages, facilitating more inclusive and generalizable advancements in synthetic speech detection.

BRSpeech-DF: A Deep Fake Synthetic Speech Dataset for Portuguese Zero-Shot TTS

Weakly supervised learning aims to reduce the cost of labeling data by using expert-designed labeling rules. However, existing methods require experts to design effective rules in a single shot, which is difficult in the absence of proper guidance and tooling. Therefore, it is still an open question whether experts should spend their limited time writing rules or instead providing instance labels via active learning. In this paper, we investigate how to exploit an expert's limited time to create effective supervision. First, to develop practical guidelines for rule creation, we conduct an exploratory analysis of diverse collections of existing expert-designed rules and find that rule precision is more important than coverage across datasets. Second, we compare rule creation to individual instance labeling via active learning and demonstrate the importance of both across 6 datasets. Third, we propose an interactive learning framework, INTERVAL, that achieves efficiency by automatically extracting candidate rules based on rich patterns (e.g., by prompting a language model), and effectiveness by soliciting expert feedback on both candidate rules and individual instances. Across 6 datasets, INTERVAL outperforms state-of-the-art weakly supervised approaches by 7% in F1. Furthermore, it requires as few as 10 queries for expert feedback to reach F1 values that existing active learning methods cannot match even with 100 queries.

Interactive Machine Teaching by Labeling Rules and Instances

Large Language Models (LLMs) fine-tuned via Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning with Verifiable Rewards (RLVR) significantly improve the alignment of human-AI values and further raise the upper bound of AI capabilities, particularly in reasoning-intensive, long-context Chain-of-Thought (long-CoT) tasks. However, existing RLHF (or RLVR) frameworks commonly face challenges such as inference bottlenecks and complexity barriers, restricting their accessibility for newcomers. To bridge this gap, we introduce \textbf{OpenRLHF}, a user-friendly, scalable, and easy-to-learn open-source RLHF framework built upon Ray, vLLM, DeepSpeed, and HuggingFace Transformers, featuring a simplified design, clear code structure, and comprehensive documentation to facilitate entry for researchers and practitioners. Experimental results show that OpenRLHF achieves superior training efficiency with speedups ranging from 1.22× to 1.68× across different model sizes compared to state-of-the-art frameworks, while requiring significantly fewer lines of code for implementation. OpenRLHF is publicly available at \url{https://github.com/OpenRLHF/OpenRLHF}, and has already been adopted by leading institutions to accelerate RLHF research and learning.

OpenRLHF: A Ray-based Easy-to-use, Scalable and High-performance RLHF Framework

Feature attribution methods, such as SHAP and LIME, explain machine learning model predictions by quantifying the influence of each input component. When applying feature attributions to explain language models, a basic question is defining the interpretable components. Traditional feature attribution methods, commonly treat individual words as atomic units. This is highly computationally inefficient for long-form text and fails to capture semantic information that spans multiple words. To address this, we present CafGa, an interactive tool for generating and evaluating feature attribution explanations at customizable granularities. CafGa supports customized segmentation with user interaction and visualizes the deletion and insertion curves for explanation assessments. Through a user study involving participants of various expertise, we confirm CafGa's usefulness, particularly among LLM practitioners. Explanations created using CafGa were also perceived as more useful compared to those generated by two fully automatic baseline methods: PartitionSHAP and MExGen, suggesting the effectiveness of the system.

CafGa: Customizing Feature Attributions to Explain Language Models

The paper presents KMatrix-2, an open-source toolkit that supports comprehensive heterogeneous knowledge collaborative enhancement for Large Language Models (LLMs). As the successor of KMatrix, our toolkit offers powerful modular components and typical enhancement patterns for convenient construction of mainstream knowledge-enhanced LLMs systems. Besides, it provides uniﬁed knowledge integration and joint knowledge retrieval methods to achieve more comprehensive heterogeneous knowledge collaborative enhancement. Compared with KMatrix which mainly focuses on descriptive knowledge, this work additionally considers procedural knowledge. Moreover, systematic inter-context and context-memory knowledge conflict resolution methods are offered for better knowledge integration. Some key research questions in heterogeneous knowledge-enhanced Large Language Models systems are analyzed, and our toolkit's capability in building such systems is validated.

KMatrix-2: A Comprehensive Heterogeneous Knowledge Collaborative Enhancement Toolkit for Large Language Model

Autoformalization serves a crucial role in connecting natural language and formal reasoning. This paper presents MASA, a novel framework for building multi-agent systems for autoformalization driven by Large Language Models (LLMs). MASA leverages collaborative agents to convert natural language statements into their formal representations. The architecture of MASA is designed with a strong emphasis on modularity, flexibility, and extensibility, allowing seamless integration of new agents and tools to adapt to a fast-evolving field. We showcase the effectiveness of MASA through use cases on real-world mathematical definitions and experiments on formal mathematics datasets. This work highlights the potential of multi-agent systems powered by the interaction of LLMs and theorem provers in enhancing the efficiency and reliability of autoformalization, providing valuable insights and support for researchers and practitioners in the field.

MASA: A Modular Framework for LLM-Driven Multi-Agent Systems for Autoformalization

Sanskrit Voyager is a web application for searching, reading, and analyzing the Sanskrit literary corpus. Unlike previous tools that require expert linguistic knowledge or manual normalization, Sanskrit Voyager enables users to search for words and phrases as they actually appear in texts, handling inflection, sandhi, and compound forms automatically while supporting any transliteration. The system integrates four core functionalities: (1) multi-dictionary lookup with morphological analysis and inflection tables; (2) real-time text parsing and annotation; (3) an interactive reader for over 900 digitalized texts; and (4) advanced corpus search with fuzzy matching and filtering. Evaluation shows over 92\% parsing accuracy on complex compounds and substantially higher search recall than BuddhaNexus on challenging queries. The application is open-source, resource-efficient, and designed for both learners and researchers, offering the first fully integrated, user-friendly platform for computational Sanskrit studies Demo video: https://www.youtube.com/watch?v=FCKlW4NKJEc

Sanskrit Voyager: Unified Web Platform for Interactive Reading and Linguistic Analysis of Sanskrit Texts

We present the iRead4Skills Intelligent Complexity Analyzer, an open-access platform specifically designed to assist educators and content developers in addressing the needs of low-literacy adults by analyzing and diagnosing text complexity. This multilingual system integrates a range of Natural Language Processing (NLP) components to assess input texts along multiple levels of granularity and linguistic dimensions in Portuguese, Spanish, and French. It assigns four tailored difficulty levels using state-of-the-art models, and introduces four diagnostic yardsticks—textual structure, lexicon, syntax, and semantics—offering users actionable feedback on specific dimensions of textual complexity. Each component of the system is supported by experiments comparing alternative models on manually annotated data.

The iRead4Skills Intelligent Complexity Analyzer

We introduce RadEval, a unified, open-source framework for evaluating radiology texts. RadEval consolidates a diverse range of metrics - from classic n‑gram overlap (BLEU, ROUGE) and contextual measures (BERTScore) to clinical concept-based scores (F1CheXbert, F1RadGraph, RaTEScore, SRR-BERT, TemporalEntityF1) and advanced LLM‑based evaluators (GREEN). We refine and standardize implementations, extend GREEN to support multiple imaging modalities with a more lightweight model, and pretrain a domain-specific radiology encoder - demonstrating strong zero-shot retrieval performance. We also release a richly annotated expert dataset with over 450 clinically significant error labels and show how different metrics correlate with radiologist judgment. Finally, RadEval provides statistical testing tools and baseline model evaluations across multiple publicly available datasets, facilitating reproducibility and robust benchmarking in radiology report generation.

RadEval: A framework for radiology text evaluation

We present LangVAE, a novel framework for modular construction of variational autoencoders (VAEs) on top of pre-trained large language models (LLMs). Such language model VAEs can encode the knowledge of their pre-trained components into more compact and semantically disentangled representations. The representations obtained in this way can be analysed with the LangVAE companion framework: LangSpace, which implements a collection of probing methods, such as vector traversal and interpolation, disentanglement measures, and cluster visualisations. LangVAE and LangSpace offer a flexible, efficient and scalable way of building and analysing textual representations, with simple integration for models available on the HuggingFace Hub. Additionally, we conducted a set of experiments with different encoder and decoder combinations, as well as annotated inputs, revealing a wide range of interactions across architectural families and sizes w.r.t. generalisation and disentanglement. Our findings demonstrate a promising framework for systematising the experimentation and understanding of textual representations.

Downloads

Next from EMNLP 2025

BRSpeech-DF: A Deep Fake Synthetic Speech Dataset for Portuguese Zero-Shot TTS

Similar lecture

InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES