China

Despite the impressive chain-of-thought(CoT) reasoning ability of large language models (LLMs), its underlying mechanisms remains unclear. In this paper, we explore the inner workings of LLM&#39;s CoT ability via the lens of neurons in the feed-forward layers. We propose an efficient method to identify reasoning-critical neurons by analyzing their activation patterns under reasoning chains of varying quality. Based on it, we devise a rather simple intervention method that directly stimulates these reasoning-critical neurons, to guide the generation of high-quality reasoning chains. Extended experiments validate the effectiveness of our method and demonstrate the critical role these identified neurons play in CoT reasoning. Our code and data will be publicly available.

EMNLP 2025

Enhancing Chain-of-Thought Reasoning via Neuron Activation Differential Analysis

and explainability

model editing

transparency

interpretability

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

ACL 2024

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

Despite the impressive chain-of-thought(CoT) reasoning ability of large language models (LLMs), its underlying mechanisms remains unclear. In this paper, we explore the inner workings of LLM's CoT ability via the lens of neurons in the feed-forward layers. We propose an efficient method to identify reasoning-critical neurons by analyzing their activation patterns under reasoning chains of varying quality. Based on it, we devise a rather simple intervention method that directly stimulates these reasoning-critical neurons, to guide the generation of high-quality reasoning chains. Extended experiments validate the effectiveness of our method and demonstrate the critical role these identified neurons play in CoT reasoning. Our code and data will be publicly available.

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) struggle with compositional generalisation, limiting their ability to systematically combine learned components to interpret novel inputs. While architectural modifications, fine-tuning, and data augmentation improve compositionality, they often have limited adaptability, face scalability constraints, or yield diminishing returns on real data. To address this, we propose CARMA, an intervention that enhances the stability and robustness of compositional reasoning in LLMs while preserving fine-tuned performance. CARMA employs mutual information regularisation and layer-wise stability constraints to mitigate feature fragmentation, ensuring structured representations persist across and within layers. We evaluate CARMA on inverse dictionary modelling and sentiment classification, measuring its impact on semantic consistency, performance stability, and robustness to lexical perturbations. Results show that CARMA reduces the variability introduced by fine-tuning, stabilises token representations, and improves compositional reasoning. While its effectiveness varies across architectures, CARMA's key strength lies in reinforcing learned structures rather than introducing new capabilities, making it a scalable auxiliary method. These findings suggest that integrating CARMA with fine-tuning can improve compositional generalisation while maintaining task-specific performance in LLMs.

CARMA: Enhanced Compositionality in LLMs via Advanced Regularisation and Mutual Information Alignment

Large Language Models (LLMs) with safe-alignment training are powerful instruments with robust language comprehension capability. Typically LLMs undergo careful alignment training involving human feedback to ensure the acceptance of safe inputs while rejection of harmful or unsafe ones. However, these humongous models are still vulnerable to jailbreak attacks, in which malicious users attempt to generate harmful outputs that safety-aligned LLMs are trained to avoid. In this study, we find that the safety mechanisms in LLMs are predominantly prevalent in the middle-to-late layers. Based on this observation, we introduce a novel white-box jailbreak method SABER (Safety Alignment Bypass via Extra Residuals) that connects two intermediate layer s and e such that s<e with a residual connection, achieving an improvement of 51% over the best performing baseline GCG on HarmBench test set. Moreover, model demonstrates only a marginal shift in perplexity when evaluated on the validation set of HarmBench.

SABER: Uncovering Vulnerabilities in Safety Alignment via Cross-Layer Residual Connection

Improvements in language model capabilities are often attributed to increasing model size or training data, but in some cases smaller models trained on curated data or with different architectural decisions can outperform larger ones trained on more tokens. What accounts for this? To quantify the impact of these design choices, we meta-analyze 92 open-source pretrained models across a wide array of scales, including state-of-the-art open-weights models as well as less performant models and those with less conventional design decisions. We find that by incorporating features besides model size and number of training tokens, we can achieve a relative 3-28% increase in ability to predict downstream performance compared with using scale alone. Analysis of model design decisions reveal insights into data composition, such as the trade-off between language and code tasks at 15-25% code, as well as the negative impact of web data on truthfulness. Broadly, our framework lays a foundation for more systematic investigation of how model development choices shape final capabilities.

Not-Just-Scaling Laws: Towards a Better Understanding of the Downstream Impact of Language Model Design Decisions

While large language models (LLMs) achieve strong performance on text-to-SQL parsing, they sometimes exhibit unexpected failures in which they are confidently incorrect. Building trustworthy text-to-SQL systems thus requires eliciting reliable uncertainty measures from the LLM. In this paper, we study the problem of providing a calibrated confidence score that conveys the likelihood of an output query being correct. Our work is the first to establish a benchmark for post-hoc calibration of LLM-based text-to-SQL parsing. In particular, we show that Platt scaling, a canonical method for calibration, provides substantial improvements over directly using raw model output probabilities as confidence scores. Furthermore, we propose a method for text-to-SQL calibration that leverages the structured nature of SQL queries to provide more granular signals of correctness, named "sub-clause frequency" (SCF) scores. Using multivariate Platt scaling (MPS), our extension of the canonical Platt scaling technique, we combine individual SCF scores into an overall accurate and calibrated score. Empirical evaluation on two popular text-to-SQL datasets shows that our approach of combining MPS and SCF yields further improvements in calibration and the related task of error detection over traditional Platt scaling.

Calibrating LLMs for Text-to-SQL Parsing by Leveraging Sub-clause Frequencies

Large language model editing methods frequently suffer from overfitting, wherein factual updates can propagate beyond their intended scope, overemphasizing the edited target even when it’s contextually inappropriate. To address this challenge, we introduce REACT (Representation Extraction And Controllable Tuning), a unified two-phase framework designed for precise and controllable knowledge editing. In the initial phase, we utilize tailored stimuli to extract latent factual representations and apply Principal Component Analysis with a simple learnbale linear transformation to compute a directional “belief shift” vector for each instance. In the second phase, we apply controllable perturbations to hidden states using the obtained vector with a magnitude scalar, gated by a pre-trained classifier that permits edits only when contextually necessary. Relevant experiments on EVOKE benchmarks demonstrate that REACT significantly reduces overfitting across nearly all evaluation metrics, and experiments on COUNTERFACT and MQuAKE shows that our method preserves balanced basic editing performance (reliability, locality, and generality) under diverse editing scenarios.

REACT: Representation Extraction And Controllable Tuning to Overcome Overfitting in LLM Knowledge Editing

Recent studies have shown that Large Language Models (LLMs) augmented with chain-of-thought (CoT) reasoning demonstrate impressive problem-solving abilities. However, in this work, we identify a recurring issue where these models occasionally generate overly short reasoning, leading to degraded performance on even simple mathematical problems. Specifically, we investigate how reasoning length is embedded in the hidden representations of reasoning models and its impact on accuracy. Our analysis reveals that reasoning length is governed by a linear direction in the representation space, allowing us to induce overly short reasoning by steering the model along this direction. Building on this insight, we introduce ThinkEdit, a simple yet effective weight-editing approach to mitigate the issue of overly short reasoning. We first identify a small subset of attention heads (approximately 4%) that predominantly drive short reasoning behavior. We then edit the output projection weights of these heads to remove the short reasoning direction. With changes to only 0.2% of the model's parameters, ThinkEdit effectively reduces overly short reasoning and yields notable accuracy gains for short reasoning outputs (+6.39%), along with an overall improvement across multiple math benchmarks (+3.34%). Our findings provide new mechanistic insights into how reasoning length is controlled within LLMs and highlight the potential of fine-grained model interventions to improve reasoning quality.

ThinkEdit: Interpretable Weight Editing to Mitigate Overly Short Thinking in Reasoning Models

Accurately evaluating the word sense disambiguation (WSD) capabilities of large language models (LLMs) remains challenging, as existing studies primarily rely on single-task evaluations and classification-based metrics that overlook the fundamental differences between generative LLMs and traditional classification models. To bridge this gap, we propose \textbf{RoDEval}, the first comprehensive evaluation framework specifically tailored for assessing LLM-based WSD methods. RoDEval introduces four novel metrics: Disambiguation Scope, Disambiguation Robustness, Disambiguation Reliability, and Definition Generation Quality Score, enabling a multifaceted evaluation of LLMs’ WSD capabilities. Experimental results using RoDEval across five mainstream LLMs uncover significant limitations in their WSD performance. Specifically, incorrect definition selections in multiple-choice WSD tasks stem not from simple neglect or forget of correct options, but rather from incomplete acquisition of the all senses for polysemous words. Instead, disambiguation reliability is often compromised by the models' persistent overconfidence. In addition, inherent biases continue to affect performance, and scaling up model parameters alone fails to meaningfully enhance their ability to generate accurate sense definitions. These findings provide actionable insights for enhancing LLMs' WSD capabilities. The source code and evaluation scripts are open-sourced at https://github.com/DayDream405/RoDEval.

RoDEval: A Robust Word Sense Disambiguation Evaluation Framework for Large Language Models

During sudden disaster events, accurately predicting public panic sentiment on social media is crucial for proactive governance and crisis management. Current efforts on this problem face three main challenges: lack of finely annotated data hinders emotion prediction studies, unmodeled risk perception causes prediction inaccuracies, and insufficient interpretability of panic formation mechanisms. We address these issues by proposing a Psychology-driven generative Agent framework (PsychoAgent) for explainable panic prediction based on emotion arousal theory. Specifically, we first construct a fine-grained open panic emotion dataset (namely COPE) via human-large language models (LLMs) collaboration to mitigate semantic bias. Then, we develop a framework integrating cross-domain heterogeneous data grounded in psychological mechanisms to model risk perception and cognitive differences in emotion generation. To enhance interpretability, we design an LLM-based role-playing agent that simulates individual psychological chains through dedicatedly designed prompts. Experimental results on our annotated dataset show that PsychoAgent improves panic emotion prediction performance by 12.6% to 21.7% compared to baseline models. Furthermore, the explainability and generalization of our approach is validated. Crucially, this represents a paradigm shift from opaque "data-driven fitting" to transparent "role-based simulation with mechanistic interpretation" for panic emotion prediction during emergencies. Our implementation is publicly available at: https://anonymous.4open.science/r/PsychoAgent-19DD.

PychoAgent: A Psychology-driven LLM Agents for Explainable Panic Prediction on Social Media during Sudden Disaster Events

Conversational aspect-based sentiment quadruple analysis (DiaASQ) is a newly-emergent task aiming to extract quadruples of target-aspect-opinion-sentiment from a conversation text. Existing studies struggle to capture complete dialogue semantics, largely due to inadequate inter-utterance modeling and the underutilization of dialogue structure. To address these issues, we propose an Inter-sentence Context Modeling and Structure-aware Representation Enhancement model (ICMSR) to extract dialogue aspect sentiment quadruple. We design the Dialog Inter-sentence Contextual Enhancer (DICE) module after the sentence-by-sentence encoding phase to enhance inter-sentence interactions and mitigate contextual fragmentation caused by traditional sequential encoding. Moreover, to fully exploit structural information within dialogues, we propose the Dialog Feature Amplifier (DFA), which consists of two submodules: STREAM and SMM. The STREAM module integrates diverse structural dialogue information to generate structure-aware sentence representations, effectively improving the modeling of intra-dialogue structural relations. Furthermore, the Structural Multi-scale Mechanism (SMM) employs a multi-scale modeling approach, simulating varying extents of contextual awareness, thereby enhancing the model’s ability to capture cross-sentence structural dependencies. We extensively evaluate our method on benchmark datasets, and the empirical results consistently confirm its effectiveness.

Inter-sentence Context Modeling and Structure-aware Representation Enhancement for Conversational Sentiment Quadruple Extraction

Recent advances in causal interpretability have extended from language models to vision-language models (VLMs), seeking to reveal their internal mechanisms through input interventions. While textual interventions often target semantics, visual interventions typically rely on coarse pixel-level perturbations, limiting semantic insights on multimodal integration. In this study, we introduce V-SEAM, a novel framework that combines Visual Semantic Editing and Attention Modulating for causal interpretation of VLMs. V-SEAM enables concept-level visual manipulations and identifies attention heads with positive or negative contributions to predictions across three semantic levels: objects, attributes, and relationships. We observe that positive heads are often shared within the same semantic level but vary across levels, while negative heads tend to generalize broadly. Finally, we introduce an automatic method to modulate key head embeddings, demonstrating enhanced performance for both LLAVA and InstructBLIP across three diverse VQA benchmarks.

Downloads

Next from EMNLP 2025

CARMA: Enhanced Compositionality in LLMs via Advanced Regularisation and Mutual Information Alignment

Similar lecture

An Investigation of Neuron Activation as a Unified Lens to Explain Chain-of-Thought Eliciting Arithmetic Reasoning of LLMs