United States

Large language models have achieved remarkable progress in text generation but still struggle with generative writing tasks. In terms of evaluation, existing evaluation benchmarks include few requirement types and writing reward models are not evaluated. In terms of training, existing studies often enhance writing ability through reinforcement learning with verifiable rewards (RLVR). Howerver, existing reward model training remains coarse-grained. To address these issues, we introduce W²Bench, a comprehensive evaluation benchmark, and WRL, a fine-grained training framework. W²Bench covers five task categories and seven requirement types, enabling systematic evaluation of both writing and writing reward models by measuring the correlation between reward rankings and golden rankings. WRL constructs positive and negative samples by dropping instruction requirements to construct positive and negative examples, allowing more precise reward model training. Experiments show that our models achieve substantial improvements on various writing benchmarks and exhibit strong generalization. We will release our code and data to support future research.

ACL 2026

 From Coarse to Fine: Benchmarking and Reward Modeling for Writing-Centric Generation Tasks

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.

poster

Reinforcement learning for Large Language Model agents is often hindered by sparse rewards in multi-step reasoning tasks. Existing approaches like Group Relative Policy Optimization treat sampled trajectories as independent chains, assigning uniform credit to all steps in each chain and ignoring the existence of critical steps that may disproportionally impact reasoning outcome. In this paper, we propose T-STAR(Tree-structured Self-Taught Agent Rectification), a framework that recovers the latent correlated reward structure across seemingly independent trajectories. Specifically, we consolidate trajectories into a unified Cognitive Tree by identifying and merging functionally similar steps/nodes. It enables an Introspective Valuation mechanism that back-propagates trajectory-level rewards through the tree to obtain a new notion of variance-reduced relative advantage at step-level. Using the Cognitive Tree, we also develop In-Context Thought Grafting to synthesize corrective reasoning by contrasting successful and failed branches at critical divergence points/steps. Our proposed Surgical Policy Optimization then capitalizes on the rich policy gradient information concentrated at these critical points/steps through a Bradley-Terry type of surgical loss. Extensive experiments across embodied, interactive, reasoning, and planning benchmarks demonstrate that T-STAR achieves consistent improvements over strong baselines, with gains most pronounced on tasks requiring extended reasoning chains.

Reason in Chains, Learn in Trees: Self-Rectification and Grafting for Multi-turn Agent Policy Optimization

Spoken conversational systems require more than accurate speech generation to have human-like conversations: to feel natural and engaging, they must produce conversational behaviour that adapts dynamically to the context. Current spoken conversational systems, however, rarely allow such customization, limiting their naturalness and usability. In this work, we present the first open, instruction-following full-duplex conversational speech model that can be trained efficiently under typical academic resource constraints. By keeping the audio encoder frozen and finetuning only the language model, our model requires just 2,000 hours of data, without relying on large-scale pretraining or multi-stage optimization. The model can follow explicit instructions to control speaker voice, conversation topic, conversational behaviour (e.g., backchanneling and interruptions), and dialogue initiation. We propose a single-stage training protocol and systematically analyze design choices. Both the model and training code is released to enable reproducible research on controllable full-duplex speech systems.

 F-Actor: Controllable Conversational Behavior in Full-Duplex Models

Multimodal sentiment analysis (MSA) in real-world scenarios is often challenged by dynamically missing modalities. Existing methods predominantly rely on deterministic imputation and rigid alignment, which compels the model to overfit noise in ambiguous regions while neglecting the decision shift induced by modality inertia. To address these issues, we propose a novel uncertainty-calibrated elastic alignment framework, termed EASE. Specifically, we employ probabilistic imputation to capture cross-modal ambiguity and leverage the estimated uncertainty to drive elastic alignment, thereby adaptively relaxing constraints in ambiguous regions to avoid rigid fitting. Meanwhile, we introduce cross-view predictive consistency constraints to unify discriminative logic across different modality views, stabilizing the decision boundary under modality degradation. Extensive experiments demonstrate that EASE consistently outperforms existing state-of-the-art baselines across multiple benchmarks, exhibiting exceptional robustness particularly under high missing-rate scenarios.

Uncertainty-Calibrated Elastic Alignment for Multimodal Sentiment Analysis with Missing Modalities

While transformer-based Large Language Models (LLMs) theoretically support massive context windows, they suffer from severe performance degradation when processing long numerical sequences. We attribute this failure to the attention dispersion in the Softmax mechanism, which prevents the model from concentrating attention. To overcome this, we propose \textbf{Sep}arate \textbf{Seq}uence (\textbf{SepSeq}), a training-free, plug-and-play framework to mitigate dispersion by strategically inserting separator tokens. Mechanistically, we demonstrate that separator tokens act as an attention anchor, recalibrating attention to focus on local segments while preserving global context. Extensive evaluations on 9 widely-adopted LLMs confirm the effectiveness of our approach: SepSeq yields an average relative accuracy improvement of 35.6% across diverse domains while reducing 16.4% inference token consumption.

SepSeq: A Training-Free Framework for Long Numerical Sequence Processing in LLMs

Large Language Models (LLMs) are increasingly tasked with creative generation, including the simulation of fictional characters. However, their ability to portray non-prosocial, antagonistic personas remains largely unexamined. We hypothesize that the safety alignment of modern LLMs creates a fundamental conflict with the task of authentically role-playing morally ambiguous or villainous characters. To investigate this, we introduce the Moral RolePlay benchmark, a new dataset featuring a four-level moral alignment scale and a balanced test set for rigorous evaluation. We task state-of-the-art LLMs with role-playing characters from moral paragons to pure villains. Our large-scale evaluation reveals a consistent, monotonic decline in role-playing fidelity as character morality decreases. We find that models struggle most with traits directly antithetical to safety principles, such as ''Deceitful'' and ''Manipulative'', often substituting nuanced malevolence with superficial aggression. Furthermore, we demonstrate that general chatbot proficiency is a poor predictor of villain role-playing ability, with highly safety-aligned models performing particularly poorly. Our work provides the first systematic evidence of this critical limitation, highlighting a key tension between model safety and creative fidelity. Our benchmark and findings pave the way for developing more nuanced, context-aware alignment methods.

Too Good to be Bad: On the Failure of LLMs to Role-Play Villains

Comprehensive software documentation is crucial yet costly to produce. Despite recent advances in large language models (LLMs), generating holistic, architecture-aware documentation at the repository level remains challenging due to complex and evolving codebases that exceed LLM context limits. Existing automated methods struggle to capture rich semantic dependencies and architectural structure. We present $\textbf{CodeWiki}$, a unified framework for automated repository-level documentation across seven mainstream programming languages. CodeWiki combines top-down hierarchical decomposition with a divide-and-conquer agent system to preserve architectural context and scale documentation generation, and a bottom-up synthesis that integrates textual descriptions with visual artifacts such as architecture and data-flow diagrams. We also introduce $\textbf{CodeWikiBench}$, a benchmark with hierarchical rubrics and LLM-based evaluation protocols. Experiments show that CodeWiki achieves a 68.79\% quality score with proprietary models, outperforming the closed-source DeepWiki baseline by 4.73\%, with especially strong gains on scripting languages. CodeWiki is released as open source to support future research.

CodeWiki: Evaluating AI’s Ability to Generate Holistic Documentation for Large-Scale Codebases

Large Language Models (LLMs) enhanced with knowledge retrieval, an approach known as Retrieval-Augmented Generation (RAG), have achieved strong performance in open-domain question answering. However, RAG remains prone to hallucinations: factually incorrect outputs may arise from inaccuracies in the model's internal knowledge and the retrieved context. Existing approaches to mitigating hallucinations often conflate factuality with faithfulness to the retrieved evidence, incorrectly labeling factually correct statements as hallucinations if they are not explicitly supported by the retrieval. In this paper, we introduce FRANQ (Faithfulness-aware Retrieval-Augmented UNcertainty Quantification), a new method for hallucination detection in RAG outputs. FRANQ applies distinct uncertainty quantification techniques to estimate factuality, conditioning on whether a statement is faithful to the retrieved context. To evaluate FRANQ and competing uncertainty quantification methods, we construct a new long-form question answering dataset annotated for both factuality and faithfulness, combining automated labeling with manual validation of challenging cases. Extensive experiments across multiple datasets, tasks, and LLMs show that FRANQ achieves more accurate detection of factual errors in RAG-generated responses compared to existing uncertainty quantification and hallucination detection approaches.

 Faithfulness-Aware Uncertainty Quantification for Fact-Checking the Output of Retrieval-Augmented Generation

In deployed large language models (LLMs), inference energy consumption has grown rapidly and has emerged as a key bottleneck in large-scale deployment, yet most existing inference efficiency methods focus on reducing FLOPs or latency, rather than explicitly modeling or enforcing end-to-end inference energy constraints. We propose EOP-LLM, an energy-oriented dynamic pruning framework that enables LLM inference under explicit per-sequence energy budgets. EOP-LLM combines a device-calibrated energy proxy with lightweight token and feed-forward (FFN) selectors, coordinated through a global dual variable, to dynamically allocate computation while preserving model quality. Extensive experiments on LLaMA 3.2 (1B/3B) and LLaMA 3.1 (8B) demonstrate that EOP-LLM consistently outperforms state-of-the-art dynamic pruning baselines under matched energy budgets, while strictly adhering to per-sequence energy constraints.

EOP-LLM: Energy Oriented Pruning for Large Language Models

A line of work in planning uses LLM not to generate a plan, but to generate a formal representation in some planning language, which can be input into a symbolic solver to deterministically find a plan. While showing improved trust and promising performance, dozens of recent publications have proposed scattered methods on a variety of benchmarks under different experimental settings. We attempt to unify the inference-time LLM-as-formalizer methodology for classical planning by proposing a unifying organizational framework based on intermediate representations. We thus systematically evaluate more than a dozen pipelines that subsume most existing work, while proposing novel ones that involve syntactically similar but high-resource intermediate languages (such as a Python wrapper of PDDL). We provide recipes for planning language generation pipelines, draw a series of conclusions showing the efficacy of their various components, and evidence their robustness against problem complexity.

Unifying Inference-Time Planning Language Generation

AI systems are fallible, and humans can make mistakes in deciding whether to trust
AI over their own judgment. Thus, improving human-AI collaboration requires that we understand when,
why, and how humans decide to rely on AI. We study two reliance decisions: delegating a task to
AI without seeing its output (whether AI is used) and evaluating AI suggestions to decide
whether to adopt them how AI output shapes final decisions).
Both matter for effective collaboration, yet prior work lacks naturalistic experiments capturing both patterns
for the same users. We address this gap by studying collaborative human--AI teams competing in a
question-answering game in which humans can choose when and how to work with AI agents to win.
Our 24 matches pair 23 expert humans with 16 AI agents, capturing 387 delegation and 1440 adoption decisions.
While human--AI collaboration performs better than either AI or humans
alone, humans make suboptimal collaboration decisions, both
under-relying on correct AI suggestions (3.7\% of opportunities missed) and over-relying when AI misleads them (1.5\%).
Both parties contribute wrong answers: reported model confidence is near chance when humans and AI disagree, 
while confirmation bias drives higher under-reliance (60.7\%) when an AI suggestion agrees with humans' initial incorrect answer.

Downloads

Next from ACL 2026

Reason in Chains, Learn in Trees: Self-Rectification and Grafting for Multi-turn Agent Policy Optimization

Similar lecture

Writing-RL: Advancing Long-form Writing via Adaptive Curriculum Reinforcement Learning