China

Embodied Question Answering (EQA) has primarily focused on indoor environments, leaving the complexities of urban settings—spanning environment, action, and perception—largely unexplored. To bridge this gap, we introduce CityEQA, a new task where an embodied agent answers open-vocabulary questions through active exploration in dynamic city spaces. To support this task, we present CityEQA-EC, the first benchmark dataset featuring 1,412 human-annotated tasks across six categories, grounded in a realistic 3D urban simulator. Moreover, we propose \Planner-Manager-Actor (PMA), a novel agent tailored for CityEQA. PMA enables long-horizon planning and hierarchical task execution: the Planner breaks down the question answering into sub-tasks, the Manager maintains an object-centric cognitive map for spatial reasoning during the process control, and the specialized Actors handle navigation, exploration, and collection sub-tasks. Experiments demonstrate that PMA achieves 60.7% of human-level answering accuracy, significantly outperforming frontier-based baselines. While promising, the performance gap compared to humans highlights the need for enhanced visual reasoning in CityEQA. This work paves the way for future advancements in urban spatial intelligence. Dataset and code are available at https://anonymous.4open.science/r/CityEQA-3027.

EMNLP 2025

CityEQA: A Hierarchical LLM Agent on Embodied Question Answering Benchmark in City Space

foundation model

embodied agent

technical paper

## 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.

In this paper, we propose the first quantitative measure for translationese---the translationese-index (T-index) for graded and generalizable measurement of translationese, computed from the likelihood ratios of two contrastively fine-tuned language models (LMs). We use a synthesized dataset and a dataset with translations in the wild to evaluate T-index's generalizability in cross-domain settings and its validity against human judgments. Our results show that T-index is both robust and efficient. T-index scored by two 0.5B LMs fine-tuned on only 1-5k pairs of synthetic data can well capture translationese in the wild. We find that the relative differences in T-indices between translations can well predict pairwise translationese annotations obtained from human annotators; and the absolute values of T-indices correlate well with human ratings of degrees of translationese (pearson r=0.568). Additionally, the correlation between T-index and existing machine translation (MT) quality estimation (QE) metrics such as BLEU and COMET is low, suggesting that T-index is not covered by these metrics and can serve as a complementary metric in MT QE. All our code and data will be released upon acceptance.

Translationese-index: Using Likelihood Ratios for Graded and Generalizable Measurement of Translationese

Large language models (LLMs) have recently demonstrated excellent performance in text embedding tasks. Previous work usually use LoRA to fine-tune existing LLMs, which are limited by the data and training gap between LLMs and embedding models. In this work, we introduce Conan-embedding-v2, an LLM-based text embedder trained from scratch. First, we add news data and multilingual pairs for LLM training to bridge the data gap. Based on this, we propose a cross-lingual retrieval dataset that enables the LLM to better integrate embeddings across different languages. Second, LLMs use causal mask with token-level loss while embedding models use bidirectional mask with sentence-level loss, this training gap makes full fine-tuning less effective than LoRA. We introduce soft-mask mechanism to gradually transition between these two types of masks, making model to learn more comprehensive representations. Based on this, we propose a dynamic hard negative mining method that exposes the model to more difficult negative examples throughout the training process. Being intuitive and effective, with only approximately 1.5B parameters, Conan-embedding-v2 achieves SOTA performance on both the Massive Text Embedding Benchmark (MTEB) and Chinese MTEB.

Conan-Embedding-v2: Training an LLM from Scratch for Text Embeddings

Information retrieval models that aim to search for documents relevant to a query have shown multiple successes, which have been applied to diverse tasks. Yet, the query from the user is oftentimes short, which challenges the retrievers to correctly fetch relevant documents. To tackle this, previous studies have proposed expanding the query with a couple of additional (user-related) features related to it. However, they may be suboptimal to effectively augment the query, and there is plenty of other information available to augment it in a relational database. Motivated by this fact, we present a novel retrieval framework called Database-Augmented Query representation (DAQu), which augments the original query with various (query-related) metadata across multiple tables. In addition, as the number of features in the metadata can be very large and there is no order among them, we encode them with the graph-based set-encoding strategy, which considers hierarchies of features in the database without order. We validate our DAQu in diverse retrieval scenarios, demonstrating that it significantly enhances overall retrieval performance over relevant baselines.

Database-Augmented Query Representation for Information Retrieval

Large Language Models (LLMs) are increasingly used in decision-making scenarios that involve risk assessment, yet their alignment with human economic rationality remains unclear. In this study, we investigate whether LLMs exhibit risk preferences consistent with human expectations across different personas. Specifically, we propose an evaluation metric called Risk Disparity Score (RDS) and assess whether LLM-generated responses reflect appropriate levels of risk aversion or risk-seeking behavior based on individual's persona. Our results reveal that while LLMs make reasonable decisions in simplified, personalized risk contexts, their performance declines in more complex economic decision-making tasks. To address this, we test whether current state-of-art alignment methods such as Direct Preference Optimization(DPO) and In Context Learning(ICL) can enhance LLM adherence to persona-specific risk preferences. We find DPO can improve the economic rationality of LLMs in loss-related parameters, offering a step toward more human-aligned AI decision-making.

Evaluating and Aligning Human Economic Risk Preferences in LLMs

Recent studies \citep{gur2018gradient,song2024does, wen2024understanding} highlight a fundamental dichotomy in deep learning optimization: Although parameter updates along the top eigendirections of the loss Hessian (Dom-space) capture most of the update magnitude, they often contribute minimally to loss reduction. In contrast, updates in the orthogonal component (Bulk-space) have smaller magnitudes but drive most learning progress. In this work, we further advance the understanding of this phenomenon and introduce the \textbf{Bulk-Space-Filtration-Accelerator (BSFA)}, a novel plug-and-play framework. BSFA accelerates training by differentially scaling update components projected onto these distinct subspaces, simultaneously enhancing stability by moderating updates in the dominant subspace and boosting convergence speed by amplifying those in the bulk-space. To ensure BSFA is both practical and scalable for contemporary large models, we introduce two key innovations: an efficient estimator using Principal Component Analysis (PCA) on historical updates for fast subspace estimation, and a block-wise strategy that applies this estimation on a per-parameter-block basis. These designs make BSFA computationally tractable and highly effective. We demonstrate BSFA's acceleration across various tasks, notably achieving approximately 2times speedup when pre-training LLaMA-72M on WikiText-103 and LLaMA-134M on OpenWebText compared to vanilla AdamW.

BSFA: Leveraging the Subspace Dichotomy to Accelerate Neural Network Training

Understanding long-form videos, such as movies and TV episodes ranging from tens of minutes to two hours, remains a major challenge for multi-modal models. Existing benchmarks often fall short in testing the full range of cognitive skills needed to process these temporally rich and narratively complex inputs. We introduce InfiniBench, a comprehensive benchmark designed to rigorously evaluate the capabilities of models in long video understanding. InfiniBench offers: (1) Over 1,000 hours of video content, with an average video length of 52.59 minutes, (2) The largest set of question-answer pairs for long video comprehension, totaling around \totalSampleNumber, (3) Eight diverse skills that span both grounding-based (e.g., scene transitions, character actions) and reasoning-based (e.g., deep context, multi-event linking) understanding, and (4) Rich annotation formats, including both multiple-choice and open-ended questions. We conduct an in-depth evaluation across both commercial (GPT-4o, Gemini 1.5 Flash) and open-source (Qwen2.5-VL, InternVL2.5) vision-language models. Results reveal that current models remain far from solving long video understanding: on grounding-based skills, the top open-source model (Qwen2.5-VL) and GPT-4o achieve only 39.4% and 48.1% accuracy, respectively. Interestingly, several models achieve non-trivial performance using only the movie or episode title, without watching the video, revealing a reliance on pre-trained world knowledge that partially compensates for the absence of visual or temporal understanding. These findings highlight critical gaps in current approaches and underscore the need for models that truly engage with long visual narratives.

InfiniBench: A Benchmark for Large Multi-Modal Models in Long-Form Movies and TV Shows

Recent advances in cross-prompt automated essay scoring (AES) typically train models jointly on all source prompts, often requiring additional access to unlabeled target prompt essays simultaneously. However, using all sources is suboptimal in our pilot study, and re-accessing source datasets during adaptation raises privacy concerns. We propose a source-free adaptation approach that selectively merges individually trained source models' parameters instead of datasets. In particular, we simulate joint training through linear combinations of task vectors—the parameter updates from fine-tuning. To optimize the combination's coefficients, we propose Prior-encoded Information Maximization (PIM), an unsupervised objective which promotes the model's score discriminability regularized by priors pre-computed from the sources. We employ Bayesian optimization as an efficient optimizer of PIM. Experimental results with LLMs on in-dataset and cross-dataset adaptation show that our method (1) consistently outperforms training jointly on all sources, (2) maintains superior robustness compared to other merging methods, (3) excels under severe distribution shifts where recent leading cross-prompt methods struggle, all while retaining computational efficiency.

Composable Cross-prompt Essay Scoring by Merging Models

Long-form writing agents require flexible integration and interaction across information retrieval, reasoning, and composition. Current approaches rely on predefined workflows and rigid thinking patterns to generate outlines before writing, resulting in constrained adaptability during writing. In this paper we propose WriteHERE, a general agent framework that achieves human-like adaptive writing through recursive task decomposition and dynamic integration of three fundamental task types: retrieval, reasoning, and composition. Our methodology features: 1) a planning mechanism that interleaves recursive task decomposition and execution, eliminating artificial restrictions on writing workflow; and 2) integration of task types that facilitates heterogeneous task decomposition. Evaluations on both fiction writing and technical report generation show that our method consistently outperforms state-of-the-art approaches across all automatic evaluation metrics, demonstrating the effectiveness and broad applicability of our proposed framework.

Beyond Outlining: Heterogeneous Recursive Planning for Adaptive Long-form Writing with Language Models

Extracting individual sentences from a document as evidence or reasoning steps is commonly done in many NLP tasks. However, extracted sentences often lack context necessary to make them understood, e.g., coreference and background information. To this end, we propose a content selection and planning framework for zero-shot decontextualisation, which determines what content should be mentioned and in what order for a sentence to be understood out of context. Specifically, given a potentially ambiguous sentence and its context, we first segment it into basic semantically-independent units. We then identify potentially ambiguous units from the given sentence, and extract relevant units from the context based on their discourse relations. Finally, we generate a content plan to rewrite the sentence by enriching each ambiguous unit with its relevant units. Experimental results demonstrate that our approach is competitive for sentence decontextualisation, producing sentences that exhibit better semantic integrity and discourse coherence, outperforming existing methods.

Improving Zero-shot Sentence Decontextualisation with Content Selection and Planning

Machine Translation (MT) systems frequently encounter gender-ambiguous occupational terms, where they must assign gender without explicit contextual cues. While individual translations in such cases may not be inherently biased, systematic patterns—such as consistently translating certain professions with specific genders—can emerge, reflecting and perpetuating societal stereotypes. This ambiguity challenges traditional instance-level single-answer evaluation approaches, as no single gold standard translation exists. To address this, we introduce GRAPE, a probability-based metric designed to evaluate gender bias by analyzing aggregated model responses. Alongside this, we present GAMBIT-MT, a benchmarking dataset in English with gender-ambiguous occupational terms. Using GRAPE, we evaluate several MT systems and examine whether their gendered translations in Greek and French align with or diverge from societal stereotypes, real-world occupational gender distributions, and normative standards.

Downloads

Next from EMNLP 2025

Translationese-index: Using Likelihood Ratios for Graded and Generalizable Measurement of Translationese