China

Reinforcement learning (RL) for large language models (LLMs) typically requires clear reward signals, which are often unavailable for open-ended (OE) questions where answer evaluation is ambiguous without scalable expert labeling. We investigate whether LLMs benefit from training on mixed data with varying reward clarity. Our approach combines Multiple-choice questions (MCQs), which offer clear binary rewards, with OE questions, for which we use simpler, potentially noisy rewards such as Jaccard similarity or LLM-based evaluators. We hypothesize that MCQs can stabilize training when mixed with OE questions. Our experiments show this mixed-data approach consistently improves medical question-answering performance across model scales.

EMNLP 2025

Training Medical QA Models Based on Mixed Rewards from Multiple-Choice and Open-Ended Questions

mixed rewards

medical llm

reinforcement learning

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

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.

Knowledge conflict often arises in retrieval-augmented generation (RAG) systems, where retrieved documents may be inconsistent with one another or contradict the model’s parametric knowledge. Existing benchmarks for investigating the phenomenon have notable limitations, including a narrow focus on the question answering setup, heavy reliance on entity substitution techniques, and a restricted range of conflict types. To address these issues, we propose a knowledge graph (KG)-based framework that generates varied and subtle conflicts between two similar yet distinct contexts, while ensuring interpretability through the explicit relational structure of KGs. Experimental results on our benchmark, MAGIC, provide intriguing insights into the inner workings of LLMs regarding knowledge conflict: both open-source and proprietary models struggle with conflict detection—especially when multi-hop reasoning is required—and often fail to pinpoint the exact source of contradictions. Finally, we present in-depth analyses that serve as a foundation for improving LLMs in integrating diverse, sometimes even conflicting, information.

MAGIC: A Multi-Hop and Graph-Based Benchmark for Inter-Context Conflicts in Retrieval-Augmented Generation

In the current landscape of large language models (LLMs), many evaluation metrics have been developed and used as rewards during training to improve specific metrics. However, balancing these metrics and dynamically adjusting reward weights remains challenging, as current approaches often fail to enhance weaker metrics. To address this, we empirically propose a Dynamic Reward Balancing Optimization framework DRBO to mitigate the "short-board effect" by measuring performance, adjusting reward weights to prioritize weaker metrics, and optimizing the model via reinforcement learning. We apply DRBO to both single-task and multi-type task scenarios, validating its effectiveness in generation with citations and online shopping conversation tasks. The results demonstrate improved overall performance and balanced optimization across multiple metrics, effectively overcoming the diversity and complexity inherent in LLMs. The code of methods and baselines is available at https://anonymous.4open.science/r/DRBO-E7D2.

DRBO: Mitigating the Bottleneck Effect via Dynamic Reward Balancing in Multi-reward LLM Optimization

Most of the existing work focuses on enabling LLMs to leverage legal rules (\eg, law articles) to tackle complex legal reasoning tasks, but ignores their ability to understand legal rules. To better evaluate the LLMs' capabilities on the task, in this work, we propose a new challenge task: Legal Paragraph Prediction (LPP), which aims to predict the legal paragraph given criminal facts. Moreover, to enhance the legal reasoning ability of LLMs, we propose a novel framework CLEAR, enabling LLMs to analyze legal cases with the guidance of legal rule insights. The CLEAR contains four key components, where the \textit{Legal Rules Retriever} aims to retrieve legal rule knowledge, and the \textit{Rule Insights Generator} is used to generate legal insights guiding the LLM's reasoning, then the \textit{Case Analyzer} analyze the case with the guidance of legal rule insights given criminal facts. Finally, the \textit{Legal Reasoner} synthesizes the criminal facts, legal rule insights, and analysis results to derive the final decision. By conducting extensive experiments on a real-world dataset, experimental results validate the effectiveness of our proposed model. Our codes and dataset are available at \url{https://anonymous.4open.science/r/CLEAR-3048}.

CLEAR: A Framework Enabling Large Language Models to Discern Confusing Legal Paragraphs

Well-designed prompts are crucial for enhancing Large language models' (LLMs) reasoning capabilities while aligning their outputs with task requirements across diverse domains. However, manually designed prompts require expertise and iterative experimentation. While existing prompt optimization methods aim to automate this process, they rely heavily on external references such as ground truth or by humans, limiting their applicability in real-world scenarios where such data is unavailable or costly to obtain. To address this, we propose Self-Supervised Prompt Optimization (SPO), a cost-efficient framework that discovers effective prompts for both closed and open-ended tasks without requiring external reference. Motivated by the observations that prompt quality manifests directly in LLM outputs and LLMs can effectively assess adherence to task requirements, we derive evaluation and optimization signals purely from output comparisons. Specifically, SPO selects superior prompts through pairwise output comparisons evaluated by an LLM evaluator, followed by an LLM optimizer that aligns outputs with task requirements. Extensive experiments demonstrate that SPO outperforms state-of-the-art prompt optimization methods, achieving comparable or superior results with significantly lower costs (e.g., 1.1% to 5.6% of existing methods) and fewer samples (e.g., three samples).

Self-Supervised Prompt Optimization

The increasing adoption of large language models (LLMs) in cloud-based services has raised significant privacy concerns, as user inputs may inadvertently expose sensitive information. Existing text anonymization and de-identification techniques, such as rule-based redaction and scrubbing, often struggle to balance privacy preservation with text naturalness and utility. In this work, we propose a zero-shot, tree-search-based iterative sentence rewriting algorithm that systematically obfuscates or deletes private information while preserving coherence, relevance, and naturalness. Our method incrementally rewrites privacy-sensitive segments through a structured search guided by a reward model, enabling dynamic exploration of the rewriting space. Experiments on privacy-sensitive datasets show that our approach significantly outperforms existing baselines, achieving a superior balance between privacy protection and utility preservation.

Zero-Shot Privacy-Aware Text Rewriting via Iterative Tree Search

Korean legal knowledge is subject to frequent temporal updates driven by societal needs and government policies. Even minor modifications to legal provisions can have significant consequences, yet continuously retraining large language models (LLMs) to incorporate such updates is resource-intensive and impractical. To address this, we propose KoLEG, an on-the-fly Korean Legal knowledge editing framework enhanced with continuous retrieval. KoLEG employs an Editing-Aware Learning Strategy and a LawEdit Retriever, which together adaptively integrate subtle linguistic nuances and continuous legislative amendments. To support this task, we construct the Korean Legislative Amendment Dataset, explicitly designed for continuous legal knowledge updates with attention to both temporal dynamics and linguistic subtleties. KoLEG outperforms existing locate-then-edit and retrieval-based editing methods, demonstrating superior effectiveness in legal knowledge editing while preserving linguistic capabilities. Furthermore, KoLEG maintains robust performance in sequential editing, improves performance on precedent application tasks, and is qualitatively validated by legal experts.

KoLEG: On-the-Fly Korean Legal Knowledge Editing with Continuous Retrieval

Multimodal Large Language Models (MLLMs) have become powerful and widely adopted in some practical applications. However, recent research has revealed their vulnerability to multimodal jailbreak attacks, whereby the model can be induced to generate harmful content, leading to safety risks. Although most MLLMs have undergone safety alignment, recent research shows that the visual modality is still vulnerable to jailbreak attacks. In our work, we discover that by using flowcharts with partially harmful information, MLLMs can be induced to provide additional harmful details. Based on this, we propose a jailbreak attack method based on auto-generated flowcharts, FC-Attack. Specifically, FC-Attack first fine-tunes a pre-trained LLM to create a step-description generator based on benign datasets. The generator is then used to produce step descriptions corresponding to a harmful query, which are transformed into flowcharts in 3 different shapes (vertical, horizontal, and S-shaped) as visual prompts. These flowcharts are then combined with a benign textual prompt to execute the jailbreak attack on MLLMs. Our evaluations on Advbench show that FC-Attack attains an attack success rate of up to 96% via images and up to 78% via videos across multiple MLLMs. Additionally, we investigate factors affecting the attack performance, including the number of steps and the font styles in the flowcharts. We also find that FC-Attack can improve the jailbreak performance from 4% to 28% in Claude-3.5 by changing the font style. To mitigate the attack, we explore several defenses and find that AdaShield can largely reduce the jailbreak performance but with the cost of utility drop.

FC-Attack: Jailbreaking Multimodal Large Language Models via Auto-Generated Flowcharts

Rapid integration of large language models (LLMs) into societal applications has intensified concerns about their alignment with universal ethical principles, as their internal value representations remain opaque despite behavioral alignment advancements. Current approaches struggle to systematically interpret how values are encoded in neural architectures, limited by datasets that prioritize superficial judgments over mechanistic analysis. We introduce ValueLocate, a mechanistic interpretability framework grounded in the Schwartz Values Survey, to address this gap. Our method first constructs ValueInsight, a dataset that operationalizes four dimensions of universal value through behavioral contexts in the real world. Leveraging this dataset, we develop a neuron identification method that calculates activation differences between opposing value aspects, enabling precise localization of value-critical neurons without relying on computationally intensive attribution methods. Our proposed validation method demonstrates that targeted manipulation of these neurons effectively alters model value orientations, establishing causal relationships between neurons and value representations. This work advances the foundation for value alignment by bridging psychological value frameworks with neuron analysis in LLMs.

Understanding How Value Neurons Shape the Generation of Specified Values in LLMs

We address the computational cost of constructing a model map, which embeds diverse language models into a common space for comparison via KL divergence. The map relies on log-likelihoods over a large text set, making the cost proportional to the number of texts. To reduce this cost, we propose a resampling method that selects important texts with weights proportional to the variance of log-likelihoods across models for each text. Our method significantly reduces the number of required texts while preserving the accuracy of KL divergence estimates. Experiments show that it achieves comparable performance to uniform sampling with about half as many texts, and also facilitates efficient incorporation of new models into an existing map. These results enable scalable and efficient construction of language model maps.

Likelihood Variance as Text Importance for Resampling Texts to Map Language Models

With the growing use of large language models (LLMs) in social, educational, and assistive contexts, understanding and controlling their personality traits has become increasingly important. In this survey, we provide a comprehensive overview of personality modeling in LLMs, covering methods ranging from rule-based systems to prompt engineering, fine-tuning, agent and retrieval techniques, as well as approaches to multimodal setups. We examine both qualitative and quantitative evaluation protocols, and identify key challenges including subjectivity, context dependence, and limited multimodal integration. We conclude by outlining open questions and future directions for building consistent, expressive, and trustworthy persona-driven LLMs.

Downloads

Next from EMNLP 2025

MAGIC: A Multi-Hop and Graph-Based Benchmark for Inter-Context Conflicts in Retrieval-Augmented Generation

Similar lecture

Enhancing Reinforcement Learning with Intrinsic Rewards from Language Model Critique