China

We study cost-efficient collaboration between strong and weak language models for repository-level code generation, where the weak model handles simpler tasks at lower cost, and the most challenging tasks are delegated to the strong model. While many works propose architectures for this task, few analyze performance relative to cost. We evaluate a broad spectrum of collaboration strategies: context-based, pipeline-based, and dynamic, on GitHub issue resolution. Our most effective collaborative strategy achieves equivalent performance to the strong model while reducing the cost by 40\%. Based on our findings, we offer actionable guidelines for choosing collaboration strategies under varying budget and performance constraints. Our results show that strong–weak collaboration substantially boosts the weak model’s performance at a fraction of the cost, pipeline and context-based methods being most efficient.

EMNLP 2025

An Empirical Study on Strong-Weak Model Collaboration for Repo-level Code Generation

llm/ai agents

llm efficiency

code generation and understanding

chain-of-thought

efficient models

retrieval-augmented generation

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s 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.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

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 2025". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

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

ACL 2025

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

poster

## 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*

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

Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, yet they often refuse to answer legitimate queries—a phenomenon known as overrefusal. Overrefusal typically stems from over-conservative safety alignment, causing models to treat many reasonable prompts as potentially risky. To systematically understand this issue, we probe and leverage the models’ safety decision boundaries to analyze and mitigate overrefusal. Our findings reveal that overrefusal is closely tied to misalignment at these boundary regions, where models struggle to distinguish subtle differences between benign and harmful content. Building on these insights, we present **RASS**, an automated framework for prompt generation and selection that strategically targets overrefusal prompts near the safety boundary. By harnessing steering vectors in the representation space, **RASS** efficiently identifies and curates boundary-aligned prompts, enabling more effective and targeted mitigation of overrefusal. This approach not only provides a more precise and interpretable view of model safety decisions but also seamlessly extends to multilingual scenarios. We have explored the safety decision boundaries of various LLMs and construct the **MORBench** evaluation set to facilitate robust assessment of model safety and helpfulness across multiple languages. Code and datasets will be released at https://anonymous.4open.science/r/RASS-80D3.

Understanding and Mitigating Overrefusal in LLMs from an Unveiling Perspective of Safety Decision Boundary

Accurate grading of rhinitis severity in nasal endoscopy relies heavily on the characterization of key secretion types, notably clear nasal discharge (CND) and purulent nasal secretion (PUS). However, both exhibit ambiguous appearance and high structural variability, posing challenges to automated grading under weak supervision. To address this, we propose Multimodal Learning for Mucus Anomaly Grading (MMAG), which integrates structured prompts with rank-aware vision-language modeling for joint detection and grading. Attribute prompts are constructed from clinical descriptors (e.g., secretion type, severity, location) and aligned with multi-level visual features via a dual-branch encoder. During inference, the model localizes mucus anomalies and maps the input image to severity-specific prompts (e.g., “moderate pus”), projecting them into a rank-aware feature space for progressive similarity scoring.Extensive evaluations on CND and PUS datasets show that our method achieves consistent gains over Baseline, improving AUC by 6.31% and 4.79%, and F1 score by 12.85% and 6.03%, respectively.This framework enables interpretable, annotation-efficient, and semantically grounded assessment of rhinitis severity based on mucus anomalies.

MMAG: Multimodal Learning for Mucus Anomaly Grading in Nasal Endoscopy via Semantic Attribute Prompting

In practical applications, multimodal data are often of low quality, with noisy modalities and missing modalities being typical forms that severely hinder model performance, robustness, and applicability. However, current studies address these issues separately. To this end, we propose a framework for multimodal affective computing that jointly addresses missing and noisy modalities to enhance model robustness in low-quality data scenarios. Specifically, we view missing modality as a special case of noisy modality, and propose a supervised attention framework. In contrast to traditional attention mechanisms that rely on main task loss to update the parameters, we design supervisory signals for the learning of attention weights, ensuring that attention mechanisms can focus on discriminative information and suppress noisy information. We further propose a ranking-based optimization strategy to compare the relative importance of different interactions by adding a ranking constraint for attention weights, avoiding training noise caused by inaccurate absolute labels. The proposed model consistently outperforms state-of-the-art baselines on multiple datasets under the settings of complete modalities, missing modalities, and noisy modalities.

Supervised Attention Mechanism for Low-quality Multimodal Data

This work presents a computational approach to analyze character development along the narrative timeline. The analysis characterizes changes in the protagonist's views and behavior and the interplay between them. We consider transcripts of Holocaust survivor testimonies as a test case, each telling the story of an individual in first-person terms. We focus on the survivor’s religious trajectory, examining the evolution of their disposition toward religious belief and practice as it is reflected in the testimony. Clustering the resulting trajectories in the dataset, we identify common sequences in the data. Our findings highlight multiple common structures of religiosity across the narratives: in terms of belief, a constant disposition is common, while for practice, most present an oscillating structure, serving as valuable material for historical and sociological research. This work demonstrates the potential of natural language processing for analyzing character evolution through thematic trajectories in narratives.

Computational Analysis of Character Development in Holocaust Testimonies

Training text embedding models under differential privacy constraints is challenging due to the high dimensionality of language data and the presence of rare, identifying linguistic features. We propose DPED (Differentially Private Embedding Distillation), a framework that leverages teacher-student distillation with multi-layer noise injection to learn high-quality embeddings while providing differential privacy guarantees. DPED trains an ensemble of teacher models on disjoint subsets of sensitive text data, then transfers their knowledge to a student model through noisy aggregation at multiple layers. A rare-word-aware strategy adaptively handles infrequent words, improving privacy-utility trade-offs. Experiments on benchmark datasets demonstrate that DPED outperforms standard differentially private training methods, achieving substantially higher utility at the same privacy budget.

DPED: Multi-Layer Noise Distillation for Privacy-Preserving Text Embeddings

Interpreting Noun-Noun Compounds remains a persistent challenge for Large Language Models (LLMs) because the semantic relation between the modifier and the head is rarely stated explicitly. Recent benchmarks frame Noun-Noun Compound Interpretation as a multiple-choice question. This setting, although prompts LLMs to yield more controlled results, still suffer from two main limitations: vague relation descriptions and failure to handle polysemous compounds. We introduce a dual-faceted textual enrichment framework that augments prompts. Description enrichment paraphrases relations into event‑oriented descriptions instantiated with the target compound to explicitly surface the hidden event connecting head and modifier. Conditioned enrichment identifies polysemous compounds leveraging qualia-role binding and assigns each compound with condition cues for disambiguation. Our method yields consistently higher accuracy across three LLM families. These gains suggest that surfacing latent compositional structure and contextual constraint is a promising path toward deeper semantic understanding in language models. The data and codebase will be made publicly available.

Enhanced Noun-Noun Compound Interpretation through Textual Enrichment

Legal citation detection in court judgments underpins reliable precedent mapping, citation analytics, and document retrieval. Extracting references to legislation and case law in the United Kingdom is especially challenging: citation styles have evolved over centuries, and judgments routinely cite foreign or historical authorities. We conduct the first systematic comparison of three modelling paradigms on this task using the Cambridge Law Corpus: (i) rule‑based regular expressions; (ii) transformer-based encoders (BERT, RoBERTa, LEGAL‑BERT, ModernBERT); and (iii) large language models (GPT‑4.1). We produced a gold‑standard high-quality corpus of 190 court judgments containing 45,179 fine-grained annotations for UK and non-UK legislation and case references. ModernBERT achieves a macro-averaged F1 of 93.3%, only marginally ahead of the other encoder-only models, yet significantly outperforming the strongest regular-expression baseline (35.42% F1) and GPT-4.1 (76.57% F1).

Detecting Legal Citations in United Kingdom Court Judgments

Retrieval-augmented language models have demonstrated performance comparable to much larger models while requiring fewer computational resources. The effectiveness of these models crucially depends on the overlap between query and retrieved context, but the optimal degree of this overlap remains unexplored. In this paper, we systematically investigate how varying levels of query--context overlap affect model performance during both training and inference. Our experiments reveal that increased overlap initially has minimal effect, but substantially improves test-time perplexity and accelerates model learning above a critical threshold. Building on these findings, we demonstrate that deliberately increasing overlap through synthetic context can enhance data efficiency and reduce training time by approximately 40% without compromising performance. We specifically generate synthetic context through paraphrasing queries. We validate our perplexity-based findings on question-answering tasks, confirming that the benefits of retrieval-augmented language modeling extend to practical applications. Our results provide empirical evidence of significant optimization potential for retrieval mechanisms in language model pretraining.

Studying the Role of Input-Neighbor Overlap in Retrieval-Augmented Language Models Training Efficiency

Inductive link prediction is emerging as a key paradigm for real-world knowledge graphs (KGs), where new entities frequently appear and models must generalize to them without retraining. Predicting links in a KG faces the challenge of guessing previously unseen entities by leveraging generalizable node features such as subgraph structure, type annotations, and ontological constraints. However, explicit type information is often lacking or incomplete. Even when available, type information in most KGs is often coarse-grained, sparse, and prone to errors due to human annotation. In this work, we explore the potential of pre-trained language models (PLMs) to enrich node representations with implicit type signals. We introduce TyleR, a Type-less yet type-awaRe approach for subgraph-based inductive link prediction that leverages PLMs for semantic enrichment. Experiments on standard benchmarks demonstrate that TyleR outperforms state-of-the-art baselines in scenarios with scarce type annotations and sparse graph connectivity. To ensure reproducibility, we share our code at https://anonymous.4open.science/r/tyler-7C2C/ .

Type-Less yet Type-Aware Inductive Link Prediction with Pretrained Language Models

The paper shows that parameter-efficient reinforcement learning (PE-RL) is a highly effective training regime to improve large language models' (LLMs) ability to answer queries on sensitive topics with a Neutral Point of View (NPOV), i.e. to provide significantly more informative, diverse and impartial answers. This is shown by evaluating PE-RL and multiple strong baselines---including LoRA finetuning (strongest baseline), SFT and RLHF. PE-RL not only improves on overall NPOV quality compared to the strongest baseline (97.06\% rightarrow 99.08\%), but also scores much higher on features linguists identify as key to separating good answers from the best answers (60.25\% rightarrow 85.21\% for presence of supportive details, 68.74\% rightarrow 91.43\% for absence of oversimplification). A qualitative analysis corroborates this. Finally, our evaluation finds no statistical differences between results on topics that appear in the training dataset and those on separated evaluation topics, which provides strong evidence that our approach to training PE-RL exhibits very effective out of topic generalization. To enable the study, and enable further future studies we also release the dataset, SHQ-NPOV, and provide a methodology to create such datasets through iterative rounds of human peer-critique and annotator training.

Downloads

Next from EMNLP 2025

Understanding and Mitigating Overrefusal in LLMs from an Unveiling Perspective of Safety Decision Boundary

Similar lecture

Synergistic Weak-Strong Collaboration by Aligning Preferences