United States

On-device fine-tuning enables privacy-preserving personalization of large language models, but mobile devices impose severe memory constraints, typically 6--12GB shared across all workloads. Existing approaches force a trade-off between exact gradients with high memory (MeBP) and low memory with noisy estimates (MeZO). We propose Memory-efficient Structured Backpropagation (MeSP), which bridges this gap by manually deriving backward passes that exploit LoRA&#39;s low-rank structure. Our key insight is that the intermediate projection $h = xA$ can be recomputed during backward at minimal cost since rank $r \ll d_{in}$, eliminating the need to store it. MeSP achieves 49\% average memory reduction compared to MeBP on Qwen2.5 models (0.5B--3B) while computing mathematically identical gradients. Our analysis also reveals that MeZO&#39;s gradient estimates show near-zero correlation with true gradients (cosine similarity $\approx$0.001), explaining its slow convergence. MeSP reduces peak memory from 361MB to 136MB for Qwen2.5-0.5B, enabling fine-tuning scenarios previously infeasible on memory-constrained devices.

ACL 2026

Memory-Efficient Structured Backpropagation for On-Device LLM Fine-Tuning

## Welcome!
"I am excited to welcome you to this year’s edition of the Conference on Empirical Methods in Natural Language Processing! Importantly, it marks the 30th edition of EMNLP. With over 8,000 submissions, more than 3,000 accepted papers, and thousands of attendees, we have come a long way from that first
workshop, which had 14 accepted papers. As the field looks ahead, Suzhou is the fitting location for celebrating this milestone: rooted in a long literary tradition, yet modern and forward-looking, and home to a large share of the NLP community." 

*Message from the General Chair, Dirk Hovy*

EMNLP 2025

Celebrate 30 Years of EMNLP! 
EMNLP 2025 will be held in Suzhou, China from November 5th to November 9th, 2025.

On-device fine-tuning enables privacy-preserving personalization of large language models, but mobile devices impose severe memory constraints, typically 6--12GB shared across all workloads. Existing approaches force a trade-off between exact gradients with high memory (MeBP) and low memory with noisy estimates (MeZO). We propose Memory-efficient Structured Backpropagation (MeSP), which bridges this gap by manually deriving backward passes that exploit LoRA's low-rank structure. Our key insight is that the intermediate projection $h = xA$ can be recomputed during backward at minimal cost since rank $r \ll d_{in}$, eliminating the need to store it. MeSP achieves 49\% average memory reduction compared to MeBP on Qwen2.5 models (0.5B--3B) while computing mathematically identical gradients. Our analysis also reveals that MeZO's gradient estimates show near-zero correlation with true gradients (cosine similarity $\approx$0.001), explaining its slow convergence. MeSP reduces peak memory from 361MB to 136MB for Qwen2.5-0.5B, enabling fine-tuning scenarios previously infeasible on memory-constrained devices.

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.

Automating scientific poster generation requires hierarchical document understanding and coherent content-layout planning.
Existing methods often rely on flat summarization or optimize content and layout separately.
As a result, they often suffer from information loss, weak logical flow, and poor visual balance.
We present PosterForest, a training-free framework for scientific poster generation.
Our method introduces the Poster Tree, a structured intermediate representation that captures document hierarchy and visual-textual semantics across multiple levels.
Building on this representation, content and layout agents perform hierarchical reasoning and recursive refinement, progressively optimizing the poster from global organization to local composition.
This joint optimization improves semantic coherence, logical flow, and visual harmony.
Experiments show that PosterForest outperforms prior methods in both automatic and human evaluations, without additional training or domain-specific supervision.

PosterForest: Hierarchical Multi-Agent Collaboration for Scientific Poster Generation

Scientific research relies on accurate information retrieval from literature to support analytical decisions.
In this work, we introduce a new task, *INformation reTRieval through literAture reVIEW* (IntraView), which aims to automate fine-grained information retrieval *faithfully* grounded in the provided content in response to research-driven queries, and propose IntrAgent, an LLM-based agent that addresses this challenging task.
In particular, IntrAgent is designed to mimic human behaviors when reading literature for information retrieval - identifying relevant sections and then iteratively extracting key details to refine the retrieved information.
It follows a two-stage pipeline: a *Section Ranking* stage that prioritizes relevant literature sections through structural-knowledge-enabled reasoning, and an *Iterative Reading* stage that continuously extracts details and synthesizes them into concise, contextually grounded answers.
To support rigorous evaluation, we introduce IntraBench, a new benchmark consisting of 315 test instances built from expert-authored questions paired with literature spanning *five* STEM domains.
Across seven backbone LLMs, IntrAgent achieves on average 13.2% higher cross-domain accuracy than state-of-the-art RAG and research-agent baselines.

 IntrAgent: An LLM Agent for Content-Grounded Information Retrieval through Literature Review

While large language models (LLMs) perform well on table tasks, they still make data referencing errors (DREs), i.e., incorrectly citing or omitting table values, despite understanding the table structure. Beyond final-answer accuracy, DREs directly compromise the correctness and reliability of intermediate reasoning steps. Yet prior studies have only offered limited, small-scale analyses. In this work, we present the first systematic evaluation of tabular data referencing errors across different models and tasks. Our results show that DREs occur across all tested models (1.7B to 20B parameters). Furthermore, we demonstrate that incorporating data referencing as a critic significantly improves answer accuracy up to 12.0\%, through critic-based filtering and rejection sampling. Finally, we trained a lightweight 4B-parameter critic model that achieves an average F1 score of 78.2\% in detecting both in-distribution and out-of-distribution DREs, and effectively assists inference for larger models.

When LLMs Read Tables Carelessly: Measuring and Reducing Data Referencing Errors

In multivariate long-term time series forecasting, it is widely believed that the effectiveness of self-attention arises from its attention matrix. We challenge this assumption with a counterintuitive finding: our experiments, conducted on three classic and three latest Transformer models, show that dot-product attention can be replaced by element-wise operations without token interaction, such as the addition and Hadamard product, while maintaining or even improving accuracy. This leads to our central hypothesis: the effectiveness of self-attention in this task stems not from the dynamic attention matrix, but from the multi-branch feature extraction inherent in the parallel projections to Query, Key, and Value matrices and their fusion. To validate this, we construct a simple multi-branch MLP that isolates the ‘multi-branch mapping with element-wise operation’ structure from the Transformer and show that it achieves competitive performance. Our results indicate that the source of performance in self-attention has been misattributed, suggesting that the true benefit lies in the architectural principle of multi-branch mapping and fusion, not in the attention matrix.

Is the Attention Matrix Really the Key to Self‑Attention in Multivariate Long‑Term Time Series Forecasting?

Audio-text retrieval systems based on Contrastive Language-Audio Pretraining (CLAP) achieve strong performance on traditional benchmarks; however, these benchmarks rely on caption-style queries that differ substantially from real-world search behavior, limiting their assessment of practical retrieval robustness. We present Omni-Embed-Audio (OEA), a retrieval-oriented encoder leveraging multimodal LLMs with native audio understanding. To systematically evaluate robustness beyond caption-style queries, we introduce User-Intent Queries (UIQs)—five formulations reflecting natural search behaviors: questions, commands, keyword tags, paraphrases, and exclusion-based negative queries. For negative queries, we develop a hard negative mining pipeline and propose discrimination metrics (HNSR, TFR) assessing models' ability to suppress acoustically similar distractors. Experiments on AudioCaps, Clotho, and MECAT show that OEA achieves comparable text-to-audio retrieval performance to state-of-the-art M2D-CLAP, while demonstrating clear advantages in two critical areas: (1) dominant text-to-text retrieval (+22% relative improvement), and (2) substantially superior hard negative discrimination (+4.3%p HNSR@10, +34.7% relative TFR@10), revealing that LLM backbones provide superior semantic understanding of complex queries.

Omni-Embed-Audio: Leveraging Multimodal LLMs for Robust Audio-Text Retrieval

Spoken language models (SpeechLMs) are increasingly real-time conversational actors. Yet many culturally consequential aspects of spoken interaction are not primarily lexical. Across sociolinguistics, linguistic anthropology, and conversation analysis, meaning emerges through how talk is produced and coordinated---prosody, timing, turn-taking, overlap, backchannels, and repair---within situated speech events. A transcript can be semantically correct yet interactionally inappropriate because many culture-bearing signals are audible and sequential rather than textual. This position paper argues for a speech-first view of cultural competence as interactional competence: the ability of a spoken agent to participate appropriately in event-situated interaction with locally normative conduct, while allowing plural acceptable realizations. Here, \emph{appropriate} does not imply generic human-likeness; in many applications, the desired behavior may instead be constrained, neutral, predictable, or tool-like under an application-specific interaction contract. We synthesize social-science foundations into a theory-derived taxonomy of culture-bearing signals in speech, identify interactional phenomena where transcript correctness fails to predict appropriateness, and ground the agenda in today's SpeechLM stacks and evaluation practice. We propose an evaluation framing that complements WER/MOS and broad capability suites by making speech events and interaction contracts explicit, diagnosing where modern pipelines lose interactional cues, and treating cultural appropriateness as a norm-conditioned target rather than generic ``naturalness.''

From Naturalness to Norms: Interactional Cultural Competence for SpeechLMs

Large reasoning models such as DeepSeek-R1 and their distilled variants achieve strong performance on complex reasoning tasks. Yet, distilling these models often demands large-scale data for supervised fine-tuning (SFT), motivating the pursuit of data-efficient training methods. To address this, we propose a skill-centric distillation framework that efficiently transfers reasoning ability to weaker models with two components: (1) Skill-based data selection, which prioritizes examples targeting the student model’s weaker skills, and (2) Skill-aware fine-tuning, which encourages explicit skill decomposition during problem solving. With only 1,000 training examples selected from a 100K teacher-generated corpus, our method surpasses random SFT baselines by +1.6% on Qwen3-4B and +1.4% on Qwen3-8B across five mathematical reasoning benchmarks. Further analysis confirms that these gains concentrate on skills emphasized during training, highlighting the effectiveness of skill-centric training for efficient reasoning distillation.

Skill-Aware Data Selection and Fine-Tuning for Data-Efficient Reasoning Distillation

Existing benchmarks for Large Language Model (LLM) agents focus on task completion under idealistic settings but overlook reliability in real-world, user-facing applications. In domains, such as in-car voice assistants, users often issue incomplete or ambiguous requests, creating intrinsic uncertainty that agents must manage through dialogue, tool use, and policy adherence. We introduce CAR-bench, a benchmark for evaluating consistency, uncertainty handling, and capability awareness in multi-turn, tool-using LLM agents in an in-car assistant domain. The environment features an LLM-simulated user, domain policies, and 58 interconnected tools spanning navigation, productivity, charging, and vehicle control. Beyond standard task completion, CAR-bench introduces Hallucination tasks that test agents’ limit-awareness under missing tools or information, and Disambiguation tasks that require resolving uncertainty through clarification or internal information gathering. Baseline results reveal large gaps between occasional and consistent success on all task types. Even frontier reasoning LLMs achieve less than 50% pass rate on Disambiguation tasks due to premature actions, and frequently violate policies or fabricate information to satisfy user requests in Hallucination tasks, underscoring the need for more reliable and self-aware LLM agents in real-world settings.

CAR-bench: Evaluating the Consistency and Limit-Awareness of LLM Agents under Real-World Uncertainty

Understanding language acquisition in language models remains an open question, yet many benchmarks focus on grammatical acceptability, with far less attention to interpreting meanings conveyed by grammatical forms.
We introduce the Linguistic Minimal-Pair Benchmark for Evaluating Constructional Understanding in Language Models (CxMP), grounded in Construction Grammar, which treats form–meaning pairings (constructions) as fundamental linguistic units.
It evaluates whether models interpret the semantic information implied by constructions, using a controlled minimal-pairs across nine types.
Our results show that constructional understanding develops more gradually and remains limited for some constructions even in large language models (LLMs), whereas performance on grammatical acceptability emerges earlier, with shallow heuristics in CxMP exhibiting a U-shaped pattern.
These findings highlight the need to broaden existing linguistic evaluations to capture meanings encoded in linguistic form.

CxMP: A Linguistic Minimal-Pair Benchmark for Evaluating Constructional Understanding in Language Models

Recent work, using the Biasing Features metric, labels a CoT as unfaithful if it omits a prompt-injected hint that affected the prediction. We argue this metric adopts a narrow notion of faithfulness and confuses unfaithfulness with incompleteness, the lossy compression needed to turn distributed transformer computation into a linear natural language narrative. On multi-hop reasoning tasks with instruct-tuned and reasoning models, many CoTs flagged as unfaithful by Biasing Features are judged faithful by other metrics, exceeding 50% in some models. With a new faithful@k metric, we show that larger inference-time budgets greatly increase hint verbalization (up to 90% in some settings), suggesting much apparent unfaithfulness is due to tight token limits. Using Causal Mediation Analysis, we further show that even non-verbalized hints can causally mediate prediction changes through the CoT. We therefore caution against relying solely on hint-based evaluations and advocate a broader interpretability toolkit, including causal mediation and corruption-based metrics. We do not claim all CoTs are faithful, only that the absence of hint words alone does not prove unfaithfulness.

Downloads

Next from ACL 2026

PosterForest: Hierarchical Multi-Agent Collaboration for Scientific Poster Generation

Similar lecture

Memory-Efficient Backpropagation for Fine-Tuning LLMs on Resource-Constrained Mobile Devices