United States

Complex agentic AI systems, powered by a coordinated ensemble of Large Language Models (LLMs), tool and memory modules, have demonstrated remarkable capabilities on intricate, multi-turn tasks. However, this success is shadowed by prohibitive economic costs and severe latency, exposing a critical, yet underexplored, trade-off. We formalize this challenge as the \textbf{Agent System Trilemma}: the inherent tension among achieving state-of-the-art performance, minimizing monetary cost, and ensuring rapid task completion. To dismantle this trilemma, we introduce EvoRoute, a self-evolving model routing paradigm that transcends static, pre-defined model assignments. Leveraging an ever-expanding knowledge base of prior experience, EvoRoute dynamically selects Pareto-optimal LLM backbones at each step, balancing accuracy, efficiency, and resource use, while continually refining its own selection policy through environment feedback. Experiments on challenging agentic benchmarks such as GAIA and BrowseComp+ demonstrate that EvoRoute, when integrated into off-the-shelf agentic systems, not only sustains or enhances system performance but also reduces execution cost by up to 80% and latency by over 70%.

ACL 2026

 EvoRoute: Experience-Driven Self-Routing LLM Agent 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.

poster

While having a significant potential for parallel processing in theory, diffusion-based non-autoregressive text generation remains inefficient due to the need for multiple denoising steps. Performance degrades sharply if a low number of steps is used, such as in flow matching. To enable accurate one-step generation, we propose a novel shortcut flow-matching model that learns to directly predict multi-step denoising outcomes in a single step. Experiments conducted on three datasets demonstrate consistent improvements over classic flow-matching, with BLEU scores more than doubling on two datasets. We also tested five different ways of extending shortcut models with commonly used techniques.

one-step Nonautoregressive Natural Language Generation with Shortcut Flow Matching Models

Large language models (LLMs) are promising for medical question answering (QA) but remain unreliable in Chinese clinical settings due to hallucinations, weak factual grounding, and difficulty handling clinically complex cases. We propose CAMEC (Complexity-Aware Multi-Expert Collaboration), a framework that combines hierarchical medical adaptation with complexity-aware expert routing for reliable Chinese medical QA. We adopt a three-stage LoRA-based supervised fine-tuning pipeline for domain adaptation, instruction following, and clinical reasoning. At inference, CAMEC routes each query by predicted complexity and selectively recruits three experts: an internal chain-of-thought (CoT) expert, a retrieval-augmented expert over a dense medical vector database, and a knowledge graph (KG) expert over a structured medical knowledge base. An LLM-as-a-Judge module evaluates and critiques expert reports, iteratively refining them into a consensus answer. Experiments on four Chinese medical benchmarks show that CAMEC consistently
outperforms strong general and medical LLM baselines, achieving 78.86\% (CMExam), 84.15\% (MedQA-CN), 78.51\% (CMMLU-Med), and 74.40\% (CMB-exam), with consistent absolute improvements over the previous state-of-the-art
HuatuoGPT-o1-7B across all benchmarks. The complexity-aware router reduces expert invocations and inference cost, 
making CAMEC both highly effective and computationally efficient.

CAMEC: Complexity-Aware Multi-Expert Collaboration for Reliable Chinese Medical Question Answering

Large Language Models (LLMs) are inherently constrained by their fixed-length context windows, which limits LLMs' ability to retain and utilize information across long-term interactions. To address this limitation, recent work has proposed external memory modules for LLMs. Using memory modules typically involves two stages: evidence retrieval and memory utilization. While prior work focuses on the architecture of memory modules and the retrieval stage, the equally critical memory utilization stage remains underexplored. Building on this, we propose MemCoRL, a two-stage alternating co-optimization reinforcement learning method. Stage 1 optimizes evidence retrieval using citation feedback and semantic accuracy from utilization as rewards. Stage 2 optimizes utilization with rewards combining semantic similarity and lexical overlap. Iterative co-optimization establishes a positive feedback loop: better retrieval improves memory utilization, which in turn refines retrieval rewards. Experimental results show our approach outperforms the leading baselines on both lexical overlap and semantic similarity metrics, confirming the co-optimization in memory retrieval and memory utilization.

MemCoRL: Alternating Co-Optimization of Memory Retrieval and Utilization via Collaborative Reinforcement Learning

Embodied planning requires agents to make coherent multi-step decisions based on dynamic visual observations and verbal goals. While recent vision-language models (VLMs) excel at static perception tasks, they struggle in interactive environments. Reinforcement learning (RL) offers a natural way to address this limitation, yet online RL approaches suffer from costly interaction and sparse rewards in embodied settings. This paper introduces ORBIT, an On-policy Reinforcement fine-tuning (RFT) framework with offline rewards for EmBodIed Task Planning, that preserves the generalization benefits of RFT while addressing the challenges of costly interaction and sparse rewards, supported by solid theoretical guarantees. Our approach is evaluated on EmbodiedBench, a recent benchmark for interactive embodied tasks, covering both in-domain and out-of-domain scenarios. Experimental results show that ORBIT achieves SOTA performance on EB-ALFRED, outperforming all closed-source and online-RL-based methods, while being substantially more efficient in training speed and computational cost, remaining robust to sub-optimal expert trajectories, and exhibiting strong generalization to unseen environments. We released all code and data at https://github.com/mail-taii/Reinforced-Reasoning-for-Embodied-Planning.

On-policy Reinforcement Fine-tuning with Offline reward for Multi-step Embodied Planning

Research on mental state reasoning in language models (LMs) has the potential to inform theories of human social cognition---such as the theory that mental state reasoning emerges in part from language exposure—and our understanding of LMs themselves. Yet much published work on LMs relies on a relatively small sample of closed-source LMs, limiting our ability to rigorously test psychological theories and evaluate LM capacities. Here, we replicate and extend published work on the false belief task by assessing LM mental state reasoning behavior across 41 open-weight models (from distinct model families). We find sensitivity to implied knowledge states in 34% of the LMs tested; however, consistent with prior work, none fully "explain away" the effect in humans. Larger LMs show increased sensitivity and also exhibit higher psychometric predictive power. Finally, we use LM behavior to generate and test a novel hypothesis about human cognition: both humans and LMs show a bias towards attributing false beliefs when knowledge states are cued using a non-factive verb ("John thinks...") than when cued indirectly ("John looks in the..."). Unlike the primary effect of knowledge states, where human sensitivity exceeds that of LMs, the magnitude of the human knowledge cue effect falls squarely within the distribution of LM effect sizes---suggesting that distributional statistics of language can in principle account for the latter but not the former in humans. These results demonstrate the value of using larger samples of open-weight LMs to test theories of human cognition and evaluate LM capacities.

Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs

The evaluation of navigation instructions remains a persistent challenge in Vision-and-Language Navigation (VLN) research. Traditional reference-based metrics such as BLEU and ROUGE fail to capture the functional utility of spatial directives, specifically whether an instruction successfully guides a navigator to the intended destination. Although existing VLN agents could serve as evaluators, their reliance on high-fidelity visual simulators introduces licensing constraints and computational costs, and perception errors further confound linguistic quality assessment. This paper introduces GROKE (Graph-based Reasoning over OSM Knowledge for instruction Evaluation), a vision-free training-free hierarchical LLM-based framework for evaluating navigation instructions using OpenStreetMap data. Through systematic ablation studies, we demonstrate that structured JSON and textual formats for spatial information substantially outperform grid-based and visual graph representations. Our hierarchical architecture combines sub-instruction planning with topological graph navigation, reducing navigation error by 68.5% compared to heuristic and sampling baselines on the Map2Seq dataset. The agent's execution success, trajectory fidelity, and decision patterns serve as proxy metrics for functional navigability given OSM-visible landmarks and topology, establishing a scalable and interpretable evaluation paradigm without visual dependencies.

GROKE: Vision-Free Navigation Instruction Evaluation via Graph Reasoning on OpenStreetMap

While recent Spoken Language Models (SLMs) have been actively deployed in real-world scenarios, they lack the capability to discern Third-Party Interruptions (TPI) from the primary user’s ongoing flow, leaving them vulnerable to contextual failures. To bridge this gap, we introduce TPI-Train, a dataset of 88K instances designed with speaker-aware hard negatives to enforce acoustic cue prioritization for interruption handling, and TPI-Bench, a comprehensive evaluation framework designed to rigorously measure the interruption-handling strategy and precise speaker discrimination in deceptive contexts. Experiments demonstrate that our dataset design mitigates semantic shortcut learning—a critical pitfall where models exploit semantic context while neglecting acoustic signals essential for discerning speaker changes. We believe our work establishes a foundational resource for overcoming text-dominated unimodal reliance in SLMs, paving the way for more robust multi-party spoken interaction. The code for the framework is publicly available at https://tpi-va.github.io

 Still Between Us? Evaluating and Improving Voice Assistant Robustness to Third-Party Interruptions

Large language models (LLMs) have demonstrated better safety performance in high-resource languages than in low-resource languages. We attribute this issue as a mismatch gap between language-agnostic semantic understanding ability and language dominant safety alignment biased toward high-resource languages. Based on above insights, we empirically identify the semantic bottleneck in LLMs: intermediate layers in which the geometry of model representations is governed primarily by shared semantic content rather than language identity. Then, we propose Language-Agnostic Semantic Alignment (LASA), which anchors safety alignment directly in semantic bottlenecks. Experiments show that LASA substantially improves safety across all languages: average attack success rate (ASR) drops from 24.7% to 2.8% on LLaMA-3.1-8B-Instruct and remains within 3–4% across Qwen2.5 and Qwen3 Instruct models (7B–32B). Besides, our analysis and method offer a representation-level perspective on LLM safety, suggesting that safety alignment requires anchoring safety understanding not in surface text, but in the model's language-agnostic semantic space.

LASA: Language-Agnostic Semantic Alignment at the Semantic Bottleneck for LLM Safety

Multilingual large language models (LLMs) are expensive to pretrain and often suffer from imbalances across languages and datasets, English-centric bias, tokenizer oversegmentation for morphologically rich low-resource languages, and the curse of multilinguality. We introduce PARAMANU, a family of Indian language-only autoregressive language models trained from scratch on open-source language-specific data for the five most spoken Indian languages: Bangla (Bengali), Hindi, Marathi, Tamil, and Telugu. All models are designed for affordability and are trained on a single GPU with a budget under $1,000, allowing under-resourced researchers to build competitive language models. To address low-resource challenges, we develop morphology-aligned, low-fertility tokenizers, and propose an interpolation-based method for token position indices in RoPE based scaling to train longer sequences efficiently. We also create instruction-tuning datasets in Bangla that are then translated to the other four languages. Despite their small size (108M-367M parameters), Paramanu achieves a strong performance-efficiency tradeoff and outperforms most larger multilingual models up to 8B across all five languages. 
The models and datasets are available at: https://huggingface.co/collections/mitodru/paramanu.

Paramanu: Compact and Competitive Monolingual Language Models for Low-Resource Morphologically Rich Indian Languages

KV cache compression promises increased throughput and efficiency with negligible loss in performance. While the gains in throughput are indisputable and recent literature has indeed shown minimal degradation on particular benchmarks, in general the consequences of compression in realistic scenarios such as multi-instruction prompting have been insufficiently studied. In this paper, we identify several pitfalls that practitioners should be aware of when deploying KV cache compressed LLMs. We evaluate five KV cache compression methods (StreamingLLM, SnapKV, TOVA, H2O, and K-Norm) on Llama3.1 8B and Qwen2.5 14B under multi-instruction prompting with IFEval. Importantly, we show that certain instructions degrade much more rapidly with compression, effectively causing them to be completely ignored by the LLM. As a practical example, we highlight system prompt leakage as a case study, empirically demonstrating the impact of compression on leakage and general instruction-following. We identify several factors that contribute to system prompt leakage: compression method, instruction order, and KV eviction bias. We then propose simple changes to KV cache eviction policies that can reduce the impact of these factors and improve the overall performance in multi-instruction tasks.

Downloads

Next from ACL 2026

one-step Nonautoregressive Natural Language Generation with Shortcut Flow Matching Models

Similar lecture

SEARL: Joint Optimization of Policy and Tool Graph Memory for Self-Evolving Agents