China

Spatial reasoning is a core component of human cognition, enabling individuals to perceive, comprehend, and interact with the physical world. It relies on a nuanced understanding of spatial structures and inter-object relationships, serving as the foundation for complex reasoning and decision-making. To investigate whether current vision-language models (VLMs) exhibit similar capability, we introduce Jigsaw-Puzzles, a novel benchmark consisting of 1,100 carefully curated real-world images with high spatial complexity. Based on this dataset, we design five tasks to rigorously evaluate VLMs’ spatial perception, structural understanding, and reasoning capabilities, while deliberately minimizing reliance on domain-specific knowledge to better isolate and assess the general spatial reasoning capability. We conduct a comprehensive evaluation across 24 state-of-the-art VLMs. The results show that even the strongest model, Gemini-2.5-Pro, achieves only 77.14% overall accuracy and performs particularly poorly on the Order Generation task, with only 30.00% accuracy, far below the 90%+ performance achieved by human participants. This persistent gap underscores the need for continued progress, positioning Jigsaw-Puzzles as a challenging and diagnostic benchmark for advancing spatial reasoning research in VLMs.

EMNLP 2025

Jigsaw-Puzzles: From Seeing to Understanding to Reasoning in Vision-Language Models

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."<br>

*Message from the General Chair, Dirk Hovy*

[**Link to Conference Handbook**](https://drive.google.com/file/d/1johU5QqVVYO4RfH7QcIORr7qrVBdzdwC/view?usp=sharing)





<br>

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 achieved remarkable success across various domains, driving significant technological advancements and innovations. Despite the rapid growth in model scale and capability, systematic, data-driven research on how structural configurations affect performance remains scarce. To address this gap, we present a large-scale dataset encompassing diverse open-source LLM structures and their performance across multiple benchmarks. Leveraging this dataset, we conduct a systematic, data mining-driven analysis to uncover the relationship between structural configurations and performance. Our study begins with a review of the historical development of LLMs and an exploration of potential future trends. We then analyze how various structural choices impact performance across benchmarks and validate our findings using mechanistic interpretability techniques. By providing data-driven insights into LLM optimization, our work aims to guide the targeted development and application of future models.

From Parameters to Performance: A Data-Driven Study on LLM Structure and Development

Large language models (LLMs) have attracted significant attention due to their impressive general capabilities across diverse downstream tasks. However, without domain-specific optimization, they often underperform on specialized knowledge benchmarks and even produce hallucination. Recent studies show that strategically infusing domain knowledge during pretraining can substantially improve downstream performance. A critical challenge lies in balancing this infusion trade-off: injecting too little domain-specific data yields insufficient specialization, whereas excessive infusion triggers catastrophic forgetting of previously acquired knowledge. In this work, we focus on the phenomenon of memory collapse induced by over-infusion. Through systematic experiments, we make two key observations, i.e. 1) Critical collapse point: each model exhibits a threshold beyond which its knowledge retention capabilities sharply degrade. 2) Scale correlation: these collapse points scale consistently with the model's size. Building on these insights, we propose a knowledge infusion scaling law that predicts the optimal amount of domain knowledge to inject into large LLMs by analyzing their smaller counterparts. Extensive experiments across different model sizes and pertaining token budgets validate both the effectiveness and generalizability of our scaling law.

How to inject knowledge efficiently? Knowledge Infusion Scaling Law for Pre-training Large Language Models

Controllable and transparent text generation has been a long-standing goal in NLP. Almost as long-standing is a general idea for addressing this challenge: Parsing text to a symbolic representation, and generating from it. However, earlier approaches were hindered by parsing and generation insufficiencies. Using modern parsers and a safety supervision mechanism, we show how close we are now to this goal. Concretely, we propose the Sentence Smith framework for English, which has three steps: 1.\ Parsing a sentence into a semantic graph. 2.\ Applying human-designed semantic manipulation rules. 3.\ Generating text from the manipulated graph. A final entailment check (4.) verifies the validity of the applied transformation. To demonstrate our framework's utility, we use it to induce hard negative text pairs that challenge text embedding models. Since the controllable generation makes it possible to clearly isolate different types of semantic shifts, we can evaluate text embedding models in a fine-grained way, also addressing an issue in current benchmarking where linguistic phenomena remain opaque. Human validation confirms that our transparent generation process produces texts of good quality.

Sentence Smith: Controllable Edits for Evaluating Text Embeddings

Though reasoning-based large language models (LLMs) have excelled in mathematics and programming, their capabilities in knowledge-intensive medical question answering remain underexplored. To address this, we introduce ReasonMed, the largest medical reasoning dataset, comprising 370k high-quality examples distilled from 1.7 million initial reasoning paths generated by various LLMs. ReasonMed is constructed through a multi-agent verification and refinement process, where we design an Error Refiner to enhance the reasoning paths by identifying and correcting error-prone steps flagged by a verifier. Leveraging ReasonMed, we systematically investigate best practices for training medical reasoning models and find that combining detailed Chain-of-Thought (CoT) reasoning with concise answer summaries yields the most effective fine-tuning strategy. Based on this strategy, we train ReasonMed-7B, which sets a new benchmark for sub-10B models, outperforming the prior best by 4.17\% and even exceeding LLaMA3.1-70B on PubMedQA by 4.60\%.

ReasonMed: A 370K Multi-Agent Generated Dataset for Advancing Medical Reasoning

Large language models (LLMs) store vast amounts of information, making them powerful yet raising privacy and safety concerns when selective knowledge removal is required. Existing unlearning strategies, ranging from gradient-based fine-tuning and model editing to sparse autoencoder (SAE) steering, either lack interpretability or fail to provide a robust defense against adversarial prompts. We propose **S**AE–Guided **S**ubspace **P**rojection **U**nlearning (**SSPU**), a novel framework that leverages SAE features to drive targeted updates in the model’s parameter space, enabling precise, interpretable, and robust unlearning. SSPU’s three-stage pipeline performs data-driven layer and feature selection, subspace construction via QR decomposition, and constrained optimization that controls activations into an “irrelevant” subspace while preserving retained knowledge. Overall, we use SAE features to construct a subspace that supervises unlearning, refining the loss and adding a regularization term to guide interpretable parameter updates. In experiments on the WMDP–Cyber forget set and three utility benchmarks (MMLU, TruthfulQA, GSM8K), SSPU reduces harmful knowledge accuracy by **3.22%** compared to the strongest baseline. It also improves adversarial robustness, lowering malicious accuracy under jailbreak prompts compared to baselines. Our findings expose the limitations of prior unlearning methods and demonstrate how interpretable subspace-guided optimization can achieve robust, controllable model behavior.

Model Unlearning via Sparse Autoencoder Subspace Guided Projections

The advancement of Argument Mining (AM) is hindered by a critical bottleneck: the scarcity of structure-annotated datasets, which are expensive to create manually. Inspired by recent successes in synthetic data generation across various NLP tasks, this paper explores methodologies for LLMs to generate synthetic data for AM. We investigate two complementary synthesis perspectives: a quality-oriented synthesis approach, which employs structure-aware paraphrasing to preserve annotation quality, and a diversity-oriented synthesis approach, which generates novel argumentative texts with diverse topics and argument structures. Experiments on three datasets show that augmenting original training data with our synthetic data, particularly when combining both quality- and diversity-oriented instances, significantly enhances the performance of existing AM models, both in full-data and low-resource settings. Moreover, the positive correlation between synthetic data volume and model performance highlights the scalability of our methods.

Exploring Quality and Diversity in Synthetic Data Generation for Argument Mining

Open Large Language Models (OLLMs) are increasingly leveraged in generative AI applications, posing new challenges for detecting their outputs. We propose OpenTuringBench, a new benchmark based on OLLMs, designed to train and evaluate machine-generated text detectors on the Turing Test and Authorship Attribution problems. OpenTuringBench focuses on a representative set of OLLMs, and features a number of challenging evaluation tasks, including human/machine-manipulated texts, out-of-domain texts, and texts from previously unseen models. We also provide OTBDetector, a contrastive learning framework to detect and attribute OLLM-based machine-generated texts. Results highlight the relevance and varying degrees of difficulty of the OpenTuringBench tasks, with our detector achieving remarkable capabilities across the various tasks and outperforming most existing detectors.

OpenTuringBench: An Open-Model-based Benchmark and Framework for Machine-Generated Text Detection and Attribution

NLP benchmarks rely on standardized datasets for training and evaluating models and are crucial for advancing the field. Traditionally, expert annotations ensure high-quality labels; however, the cost of expert annotation does not scale well with the growing demand for larger datasets required by modern models. While crowd-sourcing provides a more scalable solution, it often comes at the expense of annotation precision and consistency. Recent advancements in large language models (LLMs) offer new opportunities to enhance the annotation process, particularly for detecting label errors in existing datasets. In this work, we consider the recent approach of LLM-as-a-judge, leveraging an ensemble of LLMs to flag potentially mislabeled examples. We conduct a case study on four factual consistency datasets from the TRUE benchmark, spanning diverse NLP tasks, and on SummEval, which uses Likert-scale ratings of summary quality across multiple dimensions. We empirically analyze the labeling quality of existing datasets and compare expert, crowd-sourced, and LLM-based annotations in terms of the agreement, label quality, and efficiency, demonstrating the strengths and limitations of each annotation method. Our findings reveal a substantial number of label errors, which, when corrected, induce a significant upward shift in reported model performance. This suggests that many of the LLMs' so-called mistakes are due to label errors rather than genuine model failures. Additionally, we discuss the implications of mislabeled data and propose methods to mitigate them in training to improve performance.

Are LLMs Better than Reported? Detecting Label Errors and Mitigating Their Effect on Model Performance

Semantic similarity between two sentences depends on the aspects considered between those sentences. To study this phenomenon, Deshpande et al. (2023) proposed the Conditional Semantic Textual Similarity (C-STS) task and annotated a human-rated similarity dataset containing pairs of sentences compared under two different conditions. However, Tu et al. (2024) found various annotation issues in this dataset and showed that manually re-annotating a small portion of it leads to more accurate C-STS models. Despite these pioneering efforts, the lack of large and accurately annotated C-STS datasets remains a blocker for making progress on this task as evidenced by the subpar performance of the C-STS models. To address this training data need, we resort to Large Language Models (LLMs) to correct the condition statements and similarity ratings in the original dataset proposed by Deshpande et al. (2023). Our proposed method is able to re-annotate a large training dataset for the C-STS task with minimal manual effort. Importantly, by training a supervised C-STS model on our re-annotated training data we achieve a novel state-of-the-art (SoTA) for C-STS, thereby validating the accuracy of our dataset. The re-annotated dataset is submitted anonymously to ARR and will be publicly released upon paper acceptance to expedite the progress of C-STS research.

Annotating Training Data for Conditional Semantic Textual Similarity Measurement using Large Language Models

Speech summarization has become an essential tool for efficiently managing and accessing the growing volume of spoken and audiovisual content. However, despite its increasing importance, speech summarization remains loosely defined. The field intersects with several research areas, including speech recognition, text summarization, and specific applications like meeting summarization. This survey not only examines existing datasets and evaluation protocols, which are crucial for assessing the quality of summarization approaches, but also synthesizes recent developments in the field, highlighting the shift from traditional systems to advanced models like fine-tuned cascaded architectures and end-to-end solutions. In doing so, we surface the ongoing challenges, such as the need for realistic evaluation benchmarks, multilingual datasets, and long-context handling.

Downloads

Next from EMNLP 2025

From Parameters to Performance: A Data-Driven Study on LLM Structure and Development

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES