China

As machine translation systems approach human-level quality, traditional evaluation methodologies struggle to detect subtle translation errors. We critically examine limitations in current gold-standard approaches (MQM and ESA), including excessive categorization complexity, rough severity granularity, significant bias towards accuracy at the expense of fluency, and concerning annotation time constraints. Through in-depth analysis of English-Russian translations from WMT24, we demonstrate that employing highly qualified professional translators without strict time limitations produces substantially different results from standard evaluations. We propose RATE (Refined Assessment for Translation Evaluation) framework and collect high-quality annotations with streamlined error categorization, expanded severity ratings and multidimensional scoring that balances accuracy and fluency assessments. Our analysis reveals that state-of-the-art MT systems may have surpassed human translations in accuracy while still lagging in fluency, a critical distinction obscured by existing accuracy-biased metrics. Our findings indicate that improving evaluation depth and expertise may be as critical to advancing the field as developing better translation systems.

EMNLP 2025

Refined Assessment for Translation Evaluation: Rethinking Machine Translation Evaluation in the Era of Human-Level Systems

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.

Retrieval-Augmented Generation (RAG) has emerged as a promising technique to enhance the quality and relevance of responses generated by large language models. While recent advancements have mainly focused on improving RAG for text-based queries, RAG on multi-modal documents containing both texts and images has not been fully explored. Especially when fine-tuning does not work. This paper proposes BRIT, a novel multi-modal RAG framework that effectively unifies various text-image connections in the document into a multi-modal graph and retrieves the texts and images as a query-specific sub-graph. By traversing both image-to-text and text-to-image paths in the graph, BRIT retrieve not only directly query-relevant images and texts but also further relevant contents to answering complex cross-modal multi-hop questions. To evaluate the effectiveness of BRIT, we introduce MM-RAG test set specifically designed for multi-modal question answering tasks that require to understand the text-image relations. Our comprehensive experiments demonstrate the superiority of BRIT, highlighting its ability to handle cross-modal questions on the multi-modal documents.

BRIT: Bidirectional Retrieval over Unified Image-Text Graph

Recent advances in question answering have led to substantial progress in tasks such as multi-hop reasoning. However, global sensemaking—answering questions by synthesizing information from an entire corpus—remains a significant challenge. A prior graph-based approach to global sensemaking lacks retrieval mechanisms, topic specificity, and incurs high inference costs. To address these limitations, we propose ReTAG, a Retrieval-Enhanced, Topic-Augmented Graph framework that constructs topic-specific subgraphs and retrieves the relevant summaries for response generation. Experiments show that ReTAG improves response quality while significantly reducing inference time compared to the baseline.

ReTAG: Retrieval-Enhanced, Topic-Augmented Graph-Based Global Sensemaking

We apply definition generators based on open-weights large language models to the task of creating explanations of novel senses, taking target word usages as an input. To this end, we employ the datasets from the AXOLOTL'24 shared task on explainable semantic change modeling, which features Finnish, Russian and German languages. We fine-tune and provide publicly the open-source models performing higher than the best submissions of the aforementioned shared task, which employed closed proprietary LLMs. In addition, we find that encoder-decoder definition generators perform on par with their decoder-only counterparts.

Explaining novel senses using definition generation with open language models

LLMs have been shown to perform well in machine translation (MT) with the use of in-context learning, rivalling supervised models when translating into high-resource languages (HRLs). However, they lag behind when dealing with low-resource language (LRLs). Example selection via similarity search and supervised fine-tuning help. However the improvements they give are limited by the size, quality and diversity of existing parallel datasets. A common technique in low-resource MT is synthetic parallel data creation, the most frequent of which is backtranslation, whereby existing target-side texts are automatically translated into the source language. However, it also relies on the existence of good quality and relevant target-side texts, which are not readily available for many LRLs. In this paper, we present a new approach, TopXGen, which involves using an LLM to automatically generate topic-specific target-side data in the LRL, which can then be backtranslated to produce useful and diverse parallel texts for ICL and fine-tuning. Our intuition is that while LLMs struggle to translate into LRLs, their ability to translate well into HRLs and their multilinguality enable them to generate good quality, natural-sounding target-side texts, which can be translated well into a high-resource source language. We show that TopXGen boosts LLM translation performance during fine-tuning and in-context learning. Our code and outputs will be made freely available.

TopXGen: Topic-Diverse Parallel Data Generation for Low-Resource Machine Translation

Cultural content poses challenges for machine translation systems due to the differences in conceptualizations between cultures, where language alone may fail to convey sufficient context to capture region-specific meanings. In this work, we investigate whether images can act as cultural context in multimodal translation. We introduce \textsc{CaMMT}, a human-curated benchmark of over 5,800 triples of images along with parallel captions in English and regional languages. Using this dataset, we evaluate five Vision Language Models (VLMs) in text-only and text+image settings. Through automatic and human evaluations, we find that visual context generally improves translation quality, especially in handling Culturally-Specific Items (CSIs), disambiguation, and correct gender usage. By releasing \textsc{CaMMT}, we aim to support broader efforts in building and evaluating multimodal translation systems that are better aligned with cultural nuance and regional variation.

CaMMT: Benchmarking Culturally Aware Multimodal Machine Translation

In subjective natural language generation tasks, generating diverse perspectives is essential for fostering balanced discourse and mitigating bias. Argument generation with diverse perspectives plays a vital role in advancing the understanding of controversial claims. Despite the strong generative capabilities of large language models (LLMs), the diversity of perspectives remains insufficiently explored within argument generation task. Moreover, there remains a significant research gap in developing methods that explicitly generate multi-perspective arguments under the quality control of claim-stance alignment constraints. In this paper, we propose POEM, a textbfPerspective-aware Preference textbfOptimization with textbfEntropy textbfMaximization framework for diverse argument generation. It enhances perspective diversity through preference optimization based on the constructed preference dataset via perspective mining and diversity measuring. It further introduces entropy maximization to promote perspective diversity by encouraging dispersed semantic representations among the generated arguments. Experimental results on claim-stance argument generation benchmarks show that POEM is capable of generating diverse arguments while maintaining comparable performances in claim and stance controllability as well as text quality compared to the state-of-the-art baselines and human evaluation.

Perspective-driven Preference Optimization with Entropy Maximization for Diverse Argument Generation

Despite growing interest in Theory of Mind (ToM) tasks for evaluating language models (LMs), little is known about how LMs internally represent mental states. Understanding these internal mechanisms is critical - not only to move beyond surface-level performance, but also for model alignment and safety, where subtle misattributions of mental states may go undetected in generated outputs. In this work, we present the first systematic investigation of belief representations in LMs by probing models across different scales, training regimens, and prompts - using control tasks to rule out confounds. Our experiments provide evidence that both model size and fine‑tuning substantially improve belief representations, which are structured - not mere by-products of higher dimensionality - yet brittle: even semantically neutral prompt variations can impair them. Crucially, we show that these representations can be strengthened: targeted edits to model activations can correct wrong ToM inferences.

Brittle Minds, Fixable Activations: Understanding Belief Representations in Language Models

Combining pre-trained expert models offers substantial potential for scalable multimodal reasoning, but building a unified framework remains challenging due to the increasing diversity of input modalities and task complexity. For instance, medical diagnosis requires precise reasoning over structured clinical tables, while financial forecasting depends on interpreting plot-based data to make informed predictions. To tackle this challenge, we introduce MEXA, a training-free framework that performs modality- and task-aware aggregation of multiple expert models to enable effective multimodal reasoning across diverse and distinct domains. MEXA dynamically selects expert models based on the input modality and the task-specific reasoning demands (i.e., skills). Each expert model, specialized in a modality task pair, generates interpretable textual reasoning outputs. MEXA then aggregates and reasons over these outputs using a Large Reasoning Model (LRM) to produce the final answer. This modular design allows flexible and transparent multimodal reasoning across diverse domains without additional training overhead. We extensively evaluate our approach on diverse multimodal benchmarks, including Video Reasoning, Audio Reasoning, 3D Understanding, and Medical QA. MEXA consistently delivers performance improvements over strong multimodal baselines, highlighting the effectiveness and broad applicability of our expert-driven selection and aggregation in diverse multimodal reasoning tasks.

MEXA: Towards General Multimodal Reasoning with Dynamic Multi-Expert Aggregation

Vision-language models (VLMs) often process visual inputs through a pretrained vision encoder, followed by a projection into the language model's embedding space via a connector component. While crucial for modality fusion, the potential information loss induced by this projection step and its direct impact on model capabilities remain understudied. We propose two novel approaches to quantify such visual information loss in the projection by analyzing the latent representation space. First, we evaluate semantic information preservation by analyzing changes in k-nearest neighbor relationships between image representations, before and after projection. Second, we directly measure information loss by reconstructing visual embeddings from the projected representation, localizing loss at an image patch level. Our experiments reveal that connectors fundamentally alter visual semantic relationships---k-nearest neighbors of the visual embeddings diverge by 40-60% post-projection, correlating highly with degradation in retrieval performance. The patch-level embedding reconstruction provides interpretable insights for model behavior on visual question-answering tasks, finding that areas of high information loss reliably predict instances where models struggle.

Lost in Embeddings: Information Loss in Vision–Language Models

We investigate the use of LLM-generated data for continual pretraining of transformer encoder models in specialized domains with limited training data, using the scientific domain of invasion biology as a case study. To this end, we leverage domain-specific ontologies by enriching them with LLM-generated data and pretraining the encoder model as an ontology-informed embedding model for concept definitions. To evaluate the effectiveness of this method, we compile a benchmark specifically designed for assessing model performance in invasion biology. After demonstrating substantial improvements over standard MLM pretraining, we investigate the feasibility of applying the proposed approach to domains without comprehensive ontologies by substituting ontological concepts with concepts automatically extracted from a small corpus of scientific abstracts and establishing relationships between concepts through distributional statistics. Our results demonstrate that this automated approach achieves comparable performance using only a small set of scientific abstracts, resulting in a fully automated pipeline for enhancing domain-specific understanding of small encoder models that is especially suited for application in low-resource settings and achieves performance comparable to masked language modeling pretraining on much larger datasets.

Downloads

Next from EMNLP 2025

BRIT: Bidirectional Retrieval over Unified Image-Text Graph

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES