United States

Language models (LMs) are increasingly used in collaboration: multiple LMs trained by different parties collaborate through routing systems, multi-agent debate, model merging, and more. Critical safety risks remain in this decentralized paradigm: what if some of the models in multi-LLM systems are compromised or malicious? We first quantify the impact of malicious models by engineering four categories of malicious LMs, plug them into four types of popular model collaboration systems, and evaluate the compromised system across 10 datasets. We find that malicious models have a severe impact on the multi-LLM systems, especially for reasoning and safety domains where performance is lowered by 7.12% and 7.94% on average. We then propose mitigation strategies to alleviate the impact of malicious components, by employing external supervisors that oversee model collaboration to disable/mask them out to reduce their influence. On average, these strategies recover 95.31% of the initial performance, while making model collaboration systems fully resistant to malicious models remains an open research question.

ACL 2026

Among Us: Measuring and Mitigating Malicious Contributions in Model Collaboration Systems

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields 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.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. <br>
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) <br><br><br>
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>
<br><br><br><br>

To access this event page, you need to log in with the **email address you registered with**. <br>Access credentials will be sent to your email from Underline -  subject line "Welcome to ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

**Online Registration Form**: https://acl.swoogo.com/acl2026

Registration Required

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

technical paper

Large language models (LLMs) have shown considerable promise for annotation purposes, yet questions remain about their ability to capture human label variation (HLV) --- genuine disagreement between annotators often observed across NLP tasks. Here, we investigate how label and explanation variation manifests within and across LLMs with respect to the Natural Language Inference (NLI) task. Using zero-shot prompting with exact human annotation instructions, we treat individual model generations as participants and examine three response sampling strategies: varying generation parameters, leveraging within-family model size differences, and pooling responses from distinct LLMs. We show that, while model ensembles can generate label distributions similar to humans, they likewise exhibit distinct, idiosyncratic judgments and disagreement patterns. We further analyze explanation variation, observing that, although models generate longer explanations than humans, they demonstrate substantially less stylistic diversity. Our findings suggest that, while LLMs may serve as useful tools for generating diverse annotations, they should not be viewed as drop-in replacements for human annotators --- particularly in applications requiring authentic representation of diversity in human judgments, such as NLI.

Label and Explanation Variation in LLM-Based Annotation: a Case Study in Natural Language Inference

Real-world financial analysis involves information across multiple languages and modalities, from reports and news to scanned filings and meeting recordings. Yet most existing evaluations of LLMs in finance remain text-only, monolingual, and largely saturated by current models. To bridge these gaps, we present MultiFinBen, the first expert-annotated multilingual (five languages) and multimodal (text, vision, audio) benchmark for evaluating LLMs in realistic financial contexts. MultiFinBen introduces two new task families: multilingual financial reasoning, which tests cross-lingual evidence integration from filings and news, and financial OCR, which extracts structured text from scanned documents containing tables and charts. Rather than aggregating all available datasets, we apply a structured, difficulty-aware selection based on advanced model performance, ensuring balanced challenge and removing redundant tasks. Evaluating 21 leading LLMs shows that even frontier multimodal models like GPT-4o achieve only 46.01% overall, stronger on vision and audio but dropping sharply in multilingual settings. These findings expose persistent limitations in multilingual, multimodal, and expert-level financial reasoning. All datasets, evaluation scripts, and leaderboards are publicly released.

MultiFinBen: Benchmarking Large Language Models for Multilingual and Multimodal Financial Application

Machine unlearning aims to forget sensitive knowledge from Large Language Models (LLMs) while maintaining general utility. However, existing approaches typically treat all tokens in a response indiscriminately and enforce uncertainty over the entire vocabulary. This global treatment results in unnecessary utility degradation and extends optimization to content-agnostic regions. To address these limitations, we propose PALU (Prefix-Aware Localized Unlearning), a framework driven by a local entropy maximization objective across both temporal and vocabulary dimensions. PALU reveals that (i) suppressing the sensitive prefix alone is sufficient to sever the causal generation link, and (ii) flattening only the top-K logits is adequate to maximize uncertainty in the critical subspace. These findings allow PALU to alleviate redundant optimization across the full vocabulary and parameter space while minimizing collateral damage to general model performance. Comprehensive evaluations validate that PALU achieves superior forgetting efficacy and utility preservation compared to state-of-the-art baselines. Our code is available at https://github.com/nxZhai/PALU.

Maximizing Local Entropy Where It Matters: Prefix-Aware Localized LLM Unlearning

In multimodal information retrieval (MMIR), candidates relevant to an input query need to be retrieved from a database, where the query and database items span different modalities. As real-world databases evolve, repeatedly annotating and indexing data and re-optimizing domain-specific models across modalities is impractical. We present MULTI-SCORE, a fine-tuning-free, two-stage MMIR approach that couples efficient candidate filtering with fine-grained multimodal re-ranking. Stage-1 adopts Matryoshka representations to efficiently filter out low-relevance candidates without expensive similarity computations on full-scale representations for the entire database. Stage-2 re-ranks the filtered candidates by computing their fine-grained multimodal contextual representations with two scoring functions for semantic alignment using chain-of-thought prompting and question-answering. Experiments demonstrate state-of-the-art zero-shot retrieval on 12 MMIR tasks across 32 datasets while outperforming supervised methods on 23 datasets.

Zero-Shot Multimodal Retrieval with Multi-Scale Contextual Representations

We present AutoSchemaKG, a framework for fully autonomous knowledge graph construction that eliminates the need for predefined schemas. Our system leverages large language models to simultaneously extract knowledge triples and induce comprehensive schemas directly from text, modeling both entities and events while employing conceptualization to organize instances into semantic categories. Processing over 50 million documents, we construct ATLAS (Automated Triple Linking And Schema induction), a family of knowledge graphs with 900+ million nodes and 5.9 billion edges. This approach outperforms state-of-the-art baselines on multi-hop QA tasks and enhances LLM factuality. Notably, our schema induction achieves 92% semantic alignment with human-crafted schemas with zero manual intervention, demonstrating that billion-scale knowledge graphs with dynamically induced schemas can effectively complement parametric knowledge in large language models.

AutoSchemaKG: Autonomous Knowledge Graph Construction through Dynamic Schema Induction from Web-Scale Corpora

Embodied Action Sequence Planning focuses on the capability of embodied agents to implement action planning via environmental perception. This technology enables diverse intelligent assistance for real-world scenarios such as home and office environments. To address the limitations of existing embodied agents in meeting the requirement for proactivity and achieving joint understanding of visual and audio information, this study investigates the ability of embodied agents to proactively provide assistance through action sequence planning based on joint understanding of vision and audio perception without explicit human instructions. Correspondingly, we propose PEAP, the first multimodal proactive embodied action sequence planning dataset. We evaluate the performance of multiple Large Language Models on the PEAP dataset. The results demonstrate that these models still exhibit significant deficiencies on this task particularly lacking accurate environmental perception capabilities. Furthermore, ablation experiment and replacement experiment further corroborate that the joint understanding of multimodal information can significantly improve the models' performance on proactive embodied action sequence planning task.

PEAP: Proactive Embodied Action Sequence Planning with Joint Understanding of Vision and Audio Perception

As Large Language Models (LLMs) increasingly operate as Deep Research (DR) Agents capable of autonomous investigation and information synthesis, reliable evaluation of their task performance has become a critical bottleneck. Current benchmarks predominantly rely on static datasets, which suffer from several limitations: limited task generality, temporal misalignment, and data contamination. To address these, we introduce DR-Arena, a fully automated evaluation framework that pushes DR agents to their capability limits through dynamic investigation. DR-Arena constructs real-time Information Trees from fresh web trends to ensure the evaluation rubric is synchronized with the live world state, and employs an automated Examiner to generate structured tasks testing two orthogonal capabilities: Deep reasoning and Wide coverage. DR-Arena further adopts Adaptive Evolvement Loop, a state-machine controller that dynamically escalates task complexity based on real-time performance, demanding deeper deduction or wider aggregation until a decisive capability boundary emerges. Experiments with six advanced DR agents demonstrate that DR-Arena achieves a Spearman correlation of 0.94 with the LMSYS Search Arena leaderboard. This represents state-of-the-art alignment with human preferences without any manual efforts, validating DR-Arena as a reliable alternative for costly human adjudication.

DR-Arena: an Automated Evaluation Framework for Deep Research Agents

Spurious correlations cause deep learning models to rely on predictive shortcuts that hold in the training data but break under distribution shifts, leading to large performance drops for minority groups. Existing strategies often rely on costly group annotations or employ unstable adversarial training. In this paper, we propose Prototype-guided debiasing using Robust Invariant Feature Transformations (PRIFT), a novel framework that mitigates spurious correlations by manipulating latent space geometry. Specifically, we introduce a prototype-guided modeling approach that leverages natural language prompts to represent confounders, transforming abstract biases into interpretable geometric anchors without auxiliary classifiers. Based on these anchors, we introduce a centered projection operator that adaptively purifies representations by removing confounding deviations specific to instances while preserving essential semantic structure. Furthermore, PRIFT can handle confounding factor information at different levels, ranging from true labels to unsupervised latent inference. Experiments on four text classification benchmarks demonstrate the superiority of our method; notably, PRIFT outperforms state-of-the-art baselines and improves worst-group accuracy by over 20\% on the CivilComments dataset compared to standard empirical risk minimization.

Mitigating Spurious Correlations in Text Classification Using Latent Space Geometry

Anomaly detection aims at distinguishing between \textit{in-distribution} samples, which belong to the same distribution as the training set, and \textit{out-of-distribution} samples, which lie outside of it. In textual anomaly detection, recent approaches routinely apply anomaly detection algorithms directly to embeddings extracted from pre-trained embedding models (\textit{two-stage approaches}). However, the geometric properties of pre-trained embeddings can hinder the effectiveness of detection algorithms, which often rely on distance-based measures. In this work, we first highlight the relevance of similarity-trained models for textual anomaly detection. Beyond being trained to capture semantic similarities, these models also exhibit geometric properties that appear better suited to detection algorithms. We further demonstrate that, besides model choice, a simple post-processing step can significantly improve anomaly detection by adapting embeddings to the assumptions made by classical detection algorithms. The bulk of our experiments is done on a reformulation of the classification tasks from the MTEB benchmark into anomaly detection tasks.

Enhancing Two Steps Textual Anomaly Detection through Anisotropy Mitigation

Advances in inference methods have enabled language models to improve their predictions without additional training. These methods often prioritize raw performance over cost-effective compute usage. However, computational efficiency is key for real-world applications with resource constraints. We provide a systematic analysis of the inference scaling strategies *self-consistency*, *self-refinement*, *multi-agent debate*, and *mixture-of-agents*, to study their computational performance tradeoffs. We evaluate methods on two reasoning benchmarks (MMLU-Pro, BBH) and include extensive parameter configurations (e.g., scaling the number of parallel predictions, agents, and debate rounds) across different model sizes. Across 34 configurations and over 100 evaluations, we compute the Pareto-optimal front to select methods that achieve the best accuracy with the lowest computational budget. Notably, inference scaling improves accuracy by up to +7.1\% points over chain-of-thought at the highest evaluated budgets (20× the CoT compute budget) on MMLU-Pro. With an equal computing budget, debate and mixture-of-agents outperform self-consistency by 1.3\% and 2.7\% points, respectively. While self-consistency saturates earlier, multi-agent gains persist, particularly on more complicated tasks. We identify a simple multi-agent design guideline: mixture-of-agents is most efficient when the number of parallel generations exceeds the number of sequential aggregations.

Downloads

Next from ACL 2026

Label and Explanation Variation in LLM-Based Annotation: a Case Study in Natural Language Inference

Similar lecture

BlindGuard: Safeguarding LLM-based Multi-Agent Systems under Unknown Attacks