研究方向
基于大语言模型的语义理解与文本生成；跨模态语义理解；
AI+教育（http://www.chinese-pku.com）
科研/教育经历
2005 ~ —— 讲师/副教授，北京大学计算机学院
2003 ~ 2005， 博士后研究，北京大学
2000 ~ 2003， 博士研究生，北京大学
科研项目
[1] 国家自然科学基金面上项目: 基于大语言模型的智能作文评价关键技术研究. 2027.01-2030.12. 主持.
[2] 国家自然科学基金面上项目: 面向阅读理解的问题自动生成关键技术研究. 2021.01-2024.12. 主持.
[3] 国家自然科学基金面上项目: 基于文档的智能问答的关键技术研究与资源建设. 2018.01-2021.12. 主持.
[4] 国家自然科学基金面上项目: 基于汉语话题的句际关系自动分析研究. 2014.01-2017.12. 主持.
[5] 国家自然科学基金青年项目: 基于词语独异性特征的大规模词义标注语料库自动构建研究. 2008.01--2010.12. 主持.
[6] 国家社科基金后期资助项目: 面向语言信息处理的现代汉语并列结构研究. 2012.01-2013.06. 主持.
[7] 国家社会科学基金青年项目: 面向网络文本的词语情感义自动标注研究. 2008.06--2010.12. 主持.
[8] 国家社科基金重大项目: 面向网络文本的多视角语义分析方法、语言知识库及平台建设研究. 2013.01-2017.12，子课题负责人。
[9] 国家863项目: 面向基础教育的类人智能知识理解与推理关键技术. 2015.01-2017.12，课题骨干。
[10] 国家863项目：面向汉语语音合成的言语语义计算模型研究. 2007.07--2009.12，副组长。
[11] 人民教育出版社合作项目: 中国儿童分级阅读文本标准研制，2020.09--2021.12，主持
[12] 主持其他产学研合作课题多项，资源成果转让多项.
科研获奖
[1] 2011年，“综合型语言知识库”，国家科学技术进步二等奖.
[2] 2020年，“多情景跨领域中文文本智能校对关键技术及应用”，北京市科学技术进步二等奖.
Selected Publications in Recent Three Years（* Corresponding Author）
[1] Zhang, Z., Duan, Y... & Wu, Y*. One tool is enough: Reinforcement learning for repository-level LLM agents. ICML-2026.
[2] Chenming Tang, Yutong Yang, Kexue Wang, Yunfang Wu*. Aligning Language Models with Real-time Knowledge Editing. ACL-2026.
[3] Chenming Tang, Hsiu-Yuan Huang, Weijie Liu, Clive Bai, Saiyong Yang, Yunfang Wu*. Do not step into the same river twice: Learning to reason from trial and error. ACL-2026.
[4] Wang Cai, Yilin Wen... & Yunfang Wu*. Safety-Utility Conflicts Are Not Global: Surgical Alignment via Head-Level Diagnosis. ACL-2026.
[5] Zhengyang Wang, Sanwoo Lee, Jiaxin Wang, Chenxi Miao, Weikang Li, Yunfang Wu*. Trait-Aware Policy Optimization for Autoregressive Multi-Trait Essay Scoring. EMNLP-2026.
[6] Kun Liang, Chenming Tang, Clive Bai, Weijie Liu, Saiyong Yang, Yunfang Wu*. ADWIN: Adaptive Windows for Horizon-Aware On-Policy Distillation. EMNLP-2026.
[7] Gengyang Li, Wang Cai, Yifeng Gao, Yunfang Wu*. SyncThink: A Training-Free Strategy to Align Inference Termination with Reasoning Saturation. Findings of ACL-2026.
[8] Hsiu-Yuan Huang, Chenming Tang, Weijie Liu, Clive Bai, Saiyong Yang, Yunfang Wu*. Think outside the policy: In-context steered policy optimization. Findings of ACL-2026.
[9] Shihang Yang, Sanwoo Lee, Ningning Zhao, Yunfang Wu*. A Unified Framework to Elicit Structured Feedback for Interpretable Multi-Trait Essay Scoring. Findings of EMNLP-2026.
[10] Yutong Yang, Chenxi Miao, Weikang Li, Yunfang Wu*. ActTraitBench: Quantifying the Knowledge-Decision Gap in Large Language Models via Human-Grounded Behavioral Validation. Findings of EMNLP-2026.
[11] Kun Liang, Clive Bai, ... & Yunfang Wu*. ORBIT: On-policy Exploration-Exploitation for Controllable Multi-Budget Reasoning. Findings of EMNLP-2026.
[12] Sanwoo Lee, Kun Liang, Yunfang Wu*. Composable Cross-prompt Essay Scoring by Merging Models. EMNLP-2025.
[13] Wang Cai, Hsiuyuan Huang, Zhixiang Wang, Yunfang Wu*. Beyond Demonstrations: Dynamic Vector Construction from Latent Representations. EMNLP-2025.
[14] Zichen Wu, Hsiuyuan Huang, Yunfang Wu*. Beyond Spurious Signals: Debiasing Multimodal Large Language Models via Counterfactual Inference and Adaptive Expert Routing. Findings of EMNLP-2025.
[15] Chenming Tang, Zhixiang Wang, Hao Sun,Yunfang Wu*. Large Language Models Might Not Care What You Are Saying: Prompt Format Beats Descriptions. Findings of EMNLP-2025.
[16] Zhaoxi Zhang, Sanwoo Lee, Zhixiang Wang, Yunfang Wu*. 3DM: Distill, Dynamic Drop, and Merge for Debiasing Multi-modal Large Language Models. Findings of ACL-2025.
[17] Sanwoo Lee, Jiahao Liu, Qifan Wang, Jingang Wang, Xunliang Cai, Yunfang Wu*. Dynamic Fisher-weighted Model Merging via Bayesian Optimization. NAACL-2025.
[18] Chenming Tang, Zhixiang Wang, Yunfang Wu*. SCOI: Syntax-augmented Coverage-based In-context Example Selection for Machine Translation. EMNLP-2024.
[19] Sanwoo Lee, Yida Cai, Desong Meng, Ziyang Wang, Yunfang Wu*. Unleashing Large Language Models' Proficiency in Zero-shot Essay Scoring. Findings of EMNLP-2024.
[20] Fanyi Qu, Hao Sun, Yunfang Wu*. Unsupervised Distractor Generation via Large Language Model Distilling and Counterfactual Contrastive Decoding. Findings of ACL-2024.
[21] Chenming Tang, Fanyi Qu, Yunfang Wu*. Ungrammatical-syntax-based In-context Example Selection for Grammatical Error Correction. NAACL-2024.
[22] Ziyang Wang, Shanyu Li, HsiuYuan Huang, Yunfang Wu*. FPT: Feature Prompt Tuning for Few-shot Readability Assessment. NAACL-2024.
[23] Ming Zhang, Ke Chang, Yunfang Wu*. Multi-modal Semantic Understanding with Contrastive Cross-modal Feature Alignment. COLING-2024.
[24] Zichen Wu, HsiuYuan Huang, Fanyi Qu and Yunfang Wu*. Mixture-of-Prompt-Experts for Multi-modal Semantic Understanding. COLING-2024.
