United States

Safety alignment aims to ensure that large language models (LLMs) refuse harmful requests by post-training on harmful queries paired with refusal answers.
Although safety alignment is widely adopted in industry, the overrefusal problem where aligned LLMs also reject benign queries after safety alignment post-training, remains insufficiently studied. Such an issue degrades the usability of safety alignment in real-world applications.
In this paper, we examine how overrefusal arises under safety alignment, and propose a mitigation strategy inspired by our findings. 
We define refusal triggers as linguistic cues in the training data that elicit refusal responses, safety alignment encourages LLMs to associate refusal triggers within a training sample with refusal responses, leading aligned LLMs to refuse harmful queries.
However, the refusal triggers include not only harmful linguistic cues but also non-harmful cues, therefore causing overrefusal to benign queries.
Building on this mechanistic analysis, we propose a method that explicitly considers refusal triggers in the safety alignment fine-tuning.
Empirical results demonstrate that our approach achieves a more favorable trade-off between defense against jailbreak attacks and responsiveness to benign queries, outperforming prior methods. Warning: this paper contains harmful and biased sentences.

ACL 2026

Deactivating Refusal Triggers: Understanding and Mitigating Overrefusal in Safety Alignment

overrefusal

safety alignment

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.

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.

workshop paper

Abstractive summarization using large language models (LLMs) has become an essential tool for condensing information. Despite their ability to generate fluent summaries, these models often produce texts that are unfaithful to the original documents, manifested through hallucinations of specific words, phrases, or concepts. Current approaches to mitigating unfaithfulness typically involve post-processing corrections or contrastive learning from synthetically generated negative samples, which do not fully address the spectrum of errors that can arise in LLM-generated summaries. In this paper, we introduce a novel approach to fine-tune LLMs specifically to reduce the occurrence of unfaithful spans of text in generated summaries. We first annotate span-level hallucinations in LLM-generated summaries using automatic labeling with GPT-4. We then fine-tune the LLM using both summaries with no hallucinations and spans of hallucinated text to improve the faithfulness of the model. This paper introduces a dataset labeled to distinguish between faithful and unfaithful content and compare the performance of three techniques: gradient ascent, unlikelihood training, and task vector negation. Our experimental results show that unlikelihood training can effectively use span-level annotations to enhance summary faithfulness, reducing the number of summaries with hallucinations from 31% to 13%, a reduction of 58% on the CNN summarization dataset and from 33% to 20%, a reduction of 39% on the SAMSum dataset.

Improving the Faithfulness of LLM-based Abstractive Summarization with Span-level Unlikelihood Training

While Large Language Models (LLMs) have shown significant advancements in performance, various jailbreak attacks have posed growing safety and ethical risks. Malicious users often exploit adversarial context to deceive LLMs, prompting them to generate responses to harmful queries. In this study, we propose a new defense mechanism called Context Filtering, an input pre-processing method designed to filter out untrustworthy and unreliable context while identifying the primary prompts containing the real user intent to uncover concealed malicious intent. Given that enhancing the safety of LLMs often compromises their helpfulness, potentially affecting the experience of benign users, our method aims to improve the safety of the LLMs while preserving their original performance. We evaluate the effectiveness of our model in defending against jailbreak attacks through comparative analysis, comparing our approach with state-of-the-art defense mechanisms against six different attacks and assessing the helpfulness of LLMs under these defenses. Our model demonstrates its ability to reduce the Attack Success Rates of jailbreak attacks by up to 92% while maintaining the original LLMs' performance, achieving state-of-the-art Safety and Helpfulness balance. Notably, Context Filtering is a plug-and-play method that can be applied to all LLMs, including both white-box and black-box models, to enhance their safety without requiring any fine-tuning of the models themselves.

Context Misleads LLMs: The Role of Context Filtering in Maintaining Safe Alignment of LLMs

Multi-agent systems (MAS) are now fulfilling a broader social responsibility by serving as socio-collaborative companions, providing emotional support, facilitating collective problem-solving, and mediating peer interactions for people facing social isolation and loneliness. However, for these systems to earn and sustain user trust, the language they generate must be trustworthy: agents must maintain consistent, believable identities and coordinate in ways that mirror the organic synergy of real-world friend support groups, rather than producing generic or sycophantic responses. Current systems fall short on both counts, suffering from persona collapse, where agents revert to homogenized assistant behaviors, eroding user confidence and social sycophancy, which creates echo chambers instead of the constructive, multi-perspective dialogue that socially responsible companions demand. We propose MASCOT, a framework for trustworthy socio-collaborative conversation through a bi-level optimization strategy: 1) Persona-Aware Behavioral Alignment, an RLAIF-driven pipeline that enforces strict persona fidelity so each agent maintains a stable, predictable identity that users can rely on; and 2) Collaborative Dialogue Optimization, a meta-policy that learns group-level turn-taking dynamics by selecting the next speaker to best replicate the natural synergy of peer support groups where diverse viewpoints surface organically. Extensive evaluations across psychological support and workplace domains show that MASCOT significantly outperforms baselines, achieving up to +14.1 in Persona Consistency and +10.6 in Social Contribution, demonstrating that socially responsible multi-agent design and trustworthy language generation are both achievable and mutually reinforcing.

(This paper is a cross-submission. We are submitting this work to the TrustNLP @ ACL 2026 Workshop under the non-archival track.)

MASCOT: Towards Trustworthy Multi-Agent Socio-Collaborative Companion Systems

Most jailbreak methods achieve high attack success rates (ASR) but require attacker LLMs to craft adversarial queries and/or demand high query budgets. These resource limitations make jailbreaking expensive, and the queries generated by attacker LLMs often consist of non-interpretable random prefixes. This paper introduces Lexical Anchor Tree Search (LATS), addressing these limitations through an attacker-LLM-free method that operates purely via lexical anchor injection. LATS reformulates jailbreaking as a breadth-first tree search over multi-turn dialogues, where each node incrementally injects missing content words from the attack goal into benign prompts. Evaluations on AdvBench and HarmBench demonstrate that LATS achieves 97-100\% ASR on latest GPT, Claude, and Llama models with an average of only $\sim$6.4 queries, compared to 20+ queries required by other methods. These results highlight conversational structure as a potent and under-protected attack surface, while demonstrating superior query efficiency in an era where high ASR is readily achievable. Our code will be released to support reproducibility.

Multi-Turn Jailbreaking of Aligned LLMs via Lexical Anchor Tree Search

This paper investigates how large language models internally process nonliteral language. Analyzing five categories spanning slang, metaphor, and idioms across all 48 layers of Gemma-3-12B-IT with Gemma Scope 2 sparse autoencoders, we find a lexical familiarity gradient: processing depth depends on available prior lexical knowledge, not figurative type. Idioms diverge at L1 as entrenched units; expressions built from familiar words (metaphors, semantic-shift and constructional slang) converge at L7–9; neologisms peak at L41, activating 3× more unique features. Paraphrase-residual analysis confirms strong signals only at the gradient endpoints, yielding a three-tier hierarchy of entrenched retrieval, known-word reanalysis, and novel-word construction. We additionally identify an activation density confound in SAE feature counts that produces spurious cross-condition convergence. Overall, processing depth is better predicted by lexical familiarity than by figurative type, with implications for robustness to non-standard language and for SAE-based interpretability.

Lexical Familiarity Predicts Processing Depth for Nonliteral Language in Large Language Models

Large language models often fail to satisfy formatting instructions when they must simultaneously perform demanding tasks. We study this behavior through a prospective memory-inspired lens from cognitive psychology, using a controlled paradigm that combines verifiable formatting constraints with benchmark tasks of increasing complexity. Across three model families and over 8,000 prompts, compliance drops by 2–21% under concurrent task load. Vulnerability is highly type-dependent: terminal constraints (requiring action at the response boundary) degrade most, with drops up to 50%, while avoidance constraints remain comparatively robust. A salience-enhanced format (explicit instruction framing plus a trailing reminder) recovers much of the lost compliance, restoring performance to 90–100% in many settings. Interference is bidirectional: formatting constraints can also reduce task accuracy, with one model's GSM8K accuracy dropping from 93% to 27%. In additional stacking experiments, joint compliance declines sharply as constraints accumulate. All results use deterministic programmatic checkers, with no LLM-as-judge component, on publicly available datasets.

Did You Forget What I Asked? Prospective Memory Failures in Large Language Models

Large Vision-Language Models (LVLMs) have grown increasingly powerful in recent years, but can also exhibit harmful biases. Prior studies investigating such biases have primarily focused on demographic traits related to the visual characteristics of a person depicted in an image, such as their race or gender. This has left biases related to cultural differences (e.g., religion, socioeconomic status), which cannot be readily discerned from an individual's appearance alone, relatively understudied. To address this gap, we introduce Cultural Counterfactuals: a high-quality synthetic dataset containing nearly 60k counterfactual images for measuring cultural biases related to religion, nationality, and socioeconomic status. To ensure that cultural contexts are accurately depicted, we generate our dataset using an image-editing model to place people of different demographics into real cultural context images. This enables the construction of counterfactual image sets which depict the same person in multiple different contexts, allowing for precise measurement of the impact that cultural context differences have on LVLM outputs. We demonstrate the utility of Cultural Counterfactuals for quantifying cultural biases in popular LVLMs.

Cultural Counterfactuals: Evaluating Cultural Biases in Large Vision-Language Models with Counterfactual Examples

Large language models are increasingly used in strategic and advisory contexts, yet their safety alignment is typically evaluated in English only. We test nine models from five providers and ask whether the language of a prompt can change a model's decision in a high-stakes scenario. We use single-turn game-theoretic vignettes in which a model advises a nuclear-armed nation on whether to strike a defenseless opponent. The prompt is intentionally amoral and strategically identical across languages. We find that Japanese prompts reduce launch rates in the Claude model family: Claude Sonnet 4.6 drops from 40% to 0% in scenarios where the strike is unnecessary and from 93\% to 17\% in contested scenarios, with minimal effect when the strike is strategically rational. The effect extends to Gemini Pro 3.1 (53\% to 13\%). A cross-language experiment isolates the mechanism: when instructed to reason in Japanese in an English prompt, launch rates drop from 93\% to 37\%. It is the language the model is asked to reason in, not the language of the input, that drives the effect. When reasoning in Japanese, models spontaneously generate moral vocabulary ("moral cost", "millions of lives") that is entirely absent from the prompt. Five other models show no language effect, but they launch in nearly every condition regardless of language. The effect requires a model that already hesitates in English. These results show that LLM safety behavior is language-dependent, and that evaluating in English alone can miss both risks and safeguards encoded in other languages.

Don’t Want Your LLM to Recommend Nuclear Strike? Try Asking It in Japanese

Large language models (LLMs) are increasingly deployed with safety alignment mechanisms designed to prevent harmful outputs including hate speech, harassment, and unsafe instructions. However, existing safety evaluation frameworks remain heavily centered on English and standardized language varieties, creating a critical gap for languages characterized by extensive dialectal variation. Arabic provides a particularly important case: everyday communication across the Arab world occurs predominantly in regional dialects rather than Modern Standard Arabic (MSA), yet these dialects are systematically underrepresented in alignment training corpora and safety benchmarks. In this paper we introduce the Dialect Safety Gap, defined as systematic variation in LLM safety behavior across dialects of the same language. We argue that this phenomenon arises from the interaction between alignment training procedures and linguistic variation: safety alignment implicitly encodes normative patterns present in training datasets, and when dialectal forms diverge from those patterns, safety behavior degrades through lexical, morphological, and pragmatic mechanisms. We propose a formal framework grounded in algorithmic fairness that links dialect variation to alignment pipeline design, introduce both a binary DSG Score and a magnitude-aware Pairwise Dialect Inconsistency metric, and propose the Dialect-Aware Safety Evaluation Protocol (DASEP) as a practical evaluation framework. We demonstrate the feasibility of dialect-aware evaluation through a controlled, human-annotated prompt-probe experiment across five Arabic variety groups, revealing a structured gradient of safety degradation that correlates with linguistic distance from MSA.

Toward Dialect-Aware Safety Evaluation for Arabic Large Language Models

Modern Large Language Models (LLMs) often rely on Retrieval-Augmented Generation (RAG) to access up-to-date information; however, retrieved corpora may contain misleading, outdated, or incorrect content, raising concerns about how such evidence affects model reliability. In this work, we investigate the susceptibility of LLMs to false external evidence. Existing studies have shown that poisoned external corpora can mislead LLM responses; yet, there is still a lack of studies on the effects of different evidence properties. To bridge this gap, we design comprehensive experiments along three dimensions: styles of evidence, quantity of evidence, and the semantic similarity between external messages and the model’s internal belief. We find that instructive-style evidence demonstrates the most severe performance degradation. On the other hand, we observe a steady decline in model response quality as the amount of false evidence accumulates. Finally, we show that LLMs are more susceptible to factually incorrect evidence when their semantic similarity is close to the model’s parametric knowledge.

Downloads

Next from ACL 2026

Improving the Faithfulness of LLM-based Abstractive Summarization with Span-level Unlikelihood Training

Similar lecture

Please refuse to answer me! Mitigating Over-Refusal in Large Language Models via Adaptive Contrastive Decoding

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from ACL 2026

Improving the Faithfulness of LLM-based Abstractive Summarization with Span-level Unlikelihood Training

Similar lecture

Please refuse to answer me! Mitigating Over-Refusal in Large Language Models via Adaptive Contrastive Decoding

Downloads