United States

Reviewing medical records for clinical and insurance decisions must handle long, heterogeneous documents while producing consistent, traceable, guideline-compliant outcomes under strict latency and cost constraints. We propose GuideTree, which compiles textual guidelines into a fixed review tree of evidence-grounded verification primitives. GuideTree uses short per-document summaries only for routing each check to a minimal set of document types and candidates; final verification always reads full document text and returns structured evidence. The tree is induced offline via a cost-aware split-and-prune search and updated safely through regression-tested, versioned patches. Across 1,000 cases from four industrial review scenarios and four LLM backbones, GuideTree achieves 84.5--92.8 Macro-F1, outperforming the strongest non-expert baselines by 3.3--7.6 points and matching ExpertTree within 0.2--0.6 points (avg. 0.38). On chronic disease with Qwen3-235B-A22B-Instruct, GuideTree reduces average I/O volume to 74K input+output characters (-82% vs. long-context prompting) and average latency to 22s (-83% vs. long-context prompting), while reaching 99% decision consistency over K=5 reruns.

ACL 2026

GuideTree: Guideline-Induced Review Trees for Long Medical Records

### Welcome to ACL 2023, the 61st Annual Meeting of the Association for Computational Linguistics! 
 The conference will be held in Toronto, Canada, July 9-14, 2023. 
Following the succession of the recent conferences in our field, ACL 2023 will adopt a hybrid format.
While the impact of Covid has considerably diminished in terms of traveling, obtaining visas to Canada
entails a very long process. Moreover, the global economic conditions pose challenges for many individuals to travel to conferences. Recognizing these circumstances, we know many participants may not be
able to attend the conference in person. Therefore, we are committed to providing a great virtual platform
so everyone has the opportunity to interact with other participants and enjoy the conference. Based on the
current registered participants, approxiately 30% have chosen to attend the conference virtually. Whether
you join us in person or virtually, we sincerely hope everyone has a remarkable conference experience. 
This General Chair’s message is where I express my gratitude to the many individuals who have made
enormous contributions to the conference over the past year.

Read [**ACL 2023 General Chair's message**](https://docs.google.com/document/d/1WobYM7norbG4dI48s75HfJoD89qgX5a_F-6U8AteLSA/edit?usp=sharing/) in full.

##### **[Conference Handbook](https://2023.aclweb.org/downloads/acl2023-handbook.pdf)**

ACL 2023

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing.

technical paper

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. 
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) 
<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>

To access this event page, you need to log in with the **email address you registered with**. 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.

Despite the impressive capabilities of large language models, their substantial computational costs, latency, and privacy risks hinder their widespread deployment in real-world applications. Small Language Models (SLMs) with fewer than 10 billion parameters present a promising alternative; however, their inherent limitations in knowledge and reasoning curtail their effectiveness. Existing research primarily focuses on enhancing SLMs through scaling laws or fine-tuning strategies while overlooking the potential of using agent paradigms, such as tool use and multi-agent collaboration, to systematically compensate for the inherent weaknesses of small models. 
To address this gap, this paper presents the first large-scale, comprehensive study of <10B open-source models under three paradigms: (1) the base model, (2) a single agent equipped with tools, and (3) a multi-agent system with collaborative capabilities.
Our results show that single-agent systems achieve the best balance between performance and cost, while multi-agent setups add overhead with limited gains. Our findings highlight the importance of agent-centric design for efficient and trustworthy deployment in resource-constrained settings.

Rethinking Scale: Deployment Trade-offs of Small Language Models under Agent Paradigms

Administrative tasks in the healthcare domain share linguistic commonalities, but it can be time-consuming to manually design LLM prompts for each use case. When calling health insurers, interactive voice response (IVR) systems cause delays in patient care and increase provider burnout due to complex routing and long hold times. Thus, IVR navigation models can offer significant time savings and reduce barriers to care. We propose an automated LLM pipeline that leverages a small number of ground truth datasets to transfer specialized prompts from one task to another; specifically, we performed a cross-task transfer of our IVR navigation logic, adapting the prompt from reaching the claims department to reaching the patient benefit department. We implemented a product-quality model which obtains high success rates in large-scale industrial healthcare settings with our proposed transfer learning pipeline with only a small number of human annotations. We obtained 82\% accuracy for turn-level evaluation, surpassing an optimized human-designed prompt (79\%) while reducing the prompt complexity by up to 80\%

Transfer Learning for Generalizable Automated LLM Improvement Pipeline for IVR Navigation

Packing and shuffling tokens is a common practice in training auto-regressive language models to prevent overfitting and improve efficiency. Documents are typically concatenated to chunks of maximum sequence length (MSL) and shuffled in chunks of tokens (atom-size chunk), possibly breaking context within documents. An alternative approach is padding, which only includes one document per chunk. To optimize both packing strategies (concatenation vs padding), we explored the optimal atom size for shuffling and compared performance and efficiency. We found that in the most common setup (where average document length is greater than MSL), matching atom size to MSL yields the lowest perplexity, controlling for dataset. Also, padding yields lower final perplexity than concatenation at the cost of lower efficiency. This trade-off informs the choice of shuffling and packing methods in training LMs.

Optimizing Packing and Shuffling Strategies for Enhanced Performance in Generative Language Models

Large language models excel at technical problem solving in English but struggle when questions are posed in Bangla. While translation offers a practical solution, existing Bangla-English systems frequently mistranslate specialized terminology, altering problem semantics and degrading downstream performance. We present BanglaSTEM, a dataset of 5,000 Bangla-English sentence pairs covering computer science, mathematics, physics, chemistry, and biology. Our pipeline extracts matching passages from official bilingual curriculum textbooks using OCR, then uses LLMs to align sentences and mark technical terms. These aligned examples serve as few-shot prompts for generating over 12,000 new translation pairs from LLMs, avoiding copyright issues. Human evaluators then select the best 5,000 pairs that correctly preserve technical terminology. We also test a term-weighted BLEU metric that gives higher weight to technical words, since standard metrics treat terminology errors and common word errors equally. We show that our weighted metric correlates better with downstream accuracy in code generation and math solving, while standard BLEU gives high scores even for wrong translations. The full implementation, dataset, and model will be made publicly available.

BanglaSTEM: A Parallel Corpus and Term-Weighted Evaluation for Technical Bangla-English Translation

Aspect-Term Sentiment Analysis (ATSA) aims to predict sentiment polarity for specific aspect terms, a task complicated by conflicting sentiments and limited context in short texts. Existing graph-based approaches, including recent multi-graph methods, rely on predefined pairwise structures to capture different linguistic views. However, these approaches impose a fixed interaction granularity and depend on multiple predefined graphs, limiting their ability to adapt to sentence-specific sentiment composition and capture higher-order relationships. We propose HyperATSA, a dynamic hypergraph framework that constructs instance-specific aspect-opinion groupings directly from contextual token representations, without relying on external parsers. Hyperedges are induced via hierarchical agglomerative clustering over token embeddings, with a novel acceleration-based cutoff that identifies natural group boundaries per sentence. Experiments on Lap14, Rest14, and MAMS demonstrate consistent improvements over strong graph-based baselines, with controlled comparisons across BERT and RoBERTa backbones confirming that the gains stem from the hypergraph mechanism rather than encoder strength.

From Graphs to Hypergraphs: Enhancing Aspect-Term Sentiment Analysis via Multi-Level Relational Modeling

Faithfulness in NLP explanations is typically assessed via input perturbations, implicitly treating explanations as causal drivers of decisions, an assumption ill-suited to human judgment. In educational assessment, explanations are better understood as structured justifications accompanying holistic decisions. We propose an explainability-first framework that treats human explanations as structured artifacts independent of prediction. By modeling explanation structure explicitly, our approach enables diagnostic analyses of emphasis, consistency, and dependence. We find that scores remain invariant under evidence removal, while explanatory structure persists in tutor-specific and rubric dependent ways, revealing a mismatch between prevailing faithfulness metrics and human reasoning. This provides a model-free foundation for auditing explanations and designing NLP systems whose explanations align with human evaluative practice.

Faithfulness Beyond Plausibility: Auditing Human Explanations in Educational Assessment

Although existing frameworks for large language model (LLM) inference on CPUs are mature, they fail to fully exploit the computational potential of many-core CPU platforms. Many-core CPUs are widely deployed in web servers and high-end networking devices, and are typically organized into multiple NUMA nodes that group cores and memory. Current frameworks largely overlook the substantial overhead of cross-NUMA memory access, limiting inference scalability and intelligence enabling on such platforms. To address this limitation, we build ArcLight, a lightweight LLM inference architecture designed from the ground up for many-core CPUs. ArcLight integrates efficient memory management and thread scheduling, and introduces finely controlled tensor parallelism to mitigate the cross-node memory access wall. Experimental results show that ArcLight significantly surpasses the performance ceiling of mainstream frameworks, achieving up to 46% higher inference throughput. Moreover, ArcLight maintains compatibility with arbitrary CPU devices. ArcLight is publicly available at https://github.com/OpenBMB/ArcLight.

ArcLight: A Lightweight LLM Inference Architecture for Many-Core CPUs

Public procurement data is legally open, yet practically locked inside thousands of unstructured PDFs and inconsistent portal metadata. pAtChWoRK starts with these fragmented, unstructured sources then leverages a hybrid pipeline (traditional NLP with LLM-based technologies) to restructure this information into a navigable knowledge base. Specifically, pAtChWoRK corrects manual classification errors, extracts complex unstructured fields such as award and solvency criteria and tenders' objectives, and assists users in easily navigating the tender landscape. This unified process enables more effective handling of the transparency bottlenecks that hinder competition and oversight in public administration. A user study with practitioners across different procurement

pAtChWoRK: Patching the Pieces of Public Procurement Documents

Vision-Language Models (VLMs) have become increasingly important in AI applications, yet interpretability tools for these models lag behind those available for text-only language models. While libraries like TransformerLens have enabled significant progress in understanding language models, existing tools for VLMs are limited to basic activation probing and saving. We introduce Spectra, a library specifically designed for mechanistic interpretability of VLMs that provides unified abstractions for activation patching, attention pattern analysis, and meta-functions across diverse VLM architectures. Built on HuggingFace's Transformers, our library handles architecture-specific complexities through per-checkpoint configurations while maintaining a simple, high-level interface. We demonstrate the library's capabilities by performing interpretability experiments on a counting task, showing how researchers can easily perform experiments that were previously cumbersome to do. The library currently supports Qwen2.5-VL, Qwen3-VL, LLaVA 1.5 and SmolVLM, with a design that facilitates extension to additional architectures. The library can be found at \url{github.com/clemneo/vlm-spectra}.

Spectra: A Mechanistic Interpretability Library for Vision-Language Models

We present OpenGlass, an open-source, privacy-preserving system for low-latency multimodal assistance in real-world deployments, with a primary focus on blind and low-vision users. Deploying large multimodal models for assistive use faces a practical tension between capability and interaction: cloud or server inference can provide strong visual understanding but often incurs multi-second delays and network jitter, while wearable devices can continuously sense the environment but have limited compute and power budgets. OpenGlass resolves this gap with a split local-first design: a lightweight glasses-side sensing unit captures visual context, while a nearby consumer-grade device performs multimodal inference and streams concise spoken feedback over local Wi-Fi or cellular connections, avoiding reliance on remote cloud services. We introduce a system-level evaluation suite covering response quality, end-to-end latency from query completion to audio playback, and safety-first abstention with auditable logs. OpenGlass supports three representative interaction patterns: mobile scene guidance for obstacle avoidance, user initiated interruption of speech output, and multi-turn question answering grounded in the current scene. In real deployment on consumer-grade hardware, OpenGlass achieves sub-2s system-level end-to-end latency from query completion to first audio playback.

Downloads

Next from ACL 2026

Rethinking Scale: Deployment Trade-offs of Small Language Models under Agent Paradigms

Similar lecture

Recurrent Attention Networks for Long-text Modeling