NLPCC 2026

Workshops

The 15th CCF International Conference on Natural Language Processing and Chinese Computing

Greater Bay Area Workshop

Wednesday, 4 November 2026 · 14:00–17:00 · Auditorium

Host: Prof. Kehai Chen and Prof. Tong Zhang

TimeContentSpeaker
14:00-14:40AlayaDB.AI: The Data Foundation for Large Model EraBo Tang
14:40-15:20Token-Efficient Reasoning in LLMs: Technical Landscape and Future DirectionsXuefeng Bai
15:20-16:00Beyond Agents: From Execution to RSIChao Huang
16:00-16:15Tea break
16:15-17:00Towards Reliable Language and Multimodal Models: Knowing, Retrieving, and GroundingHou Pong (Ken) Chan

AlayaDB.AI: The Data Foundation for Large Model Era

Bo Tang

Abstract

In the large model era, vectors—serving as metadata for large models—are being applied across various aspects of these models, such as Retrieval-Augmented Generation (RAG), sparse attention, and KV caching. Our team has taken vectors as the entry point to develop an ultra-efficient LLM inference infrastructure. In this talk, I will introduce two research work: AlayaLite and AlayaJet, and briefly discuss the key challenges in building AlayaDB.AI, the data foundation for the large model era.

Biography

Dr. Bo Tang is a tenured associate professor at Southern University of Science and Technology. His research interests are big data/large model systems. He always published at top-tier conference and journals (e.g., SIGMOD, PVLDB, TODS). His research outputs have been widely used in leading IT companies (e.g., Microsoft, Tencent, and Huawei). He won Huawei Spark Award (3 times). He is awarded SIGMOD China Rising Star at 2021 and sponsored by NSFC excellent young researcher grant at 2024.

Token-Efficient Reasoning in LLMs: Technical Landscape and Future Directions

Xuefeng Bai

Abstract

Reasoning-oriented large language models (LLMs) have recently demonstrated striking capabilities on complex tasks, yet their inference suffers from a fundamental bottleneck: reasoning is inefficient and expensive. Efficient reasoning has therefore become the decisive step that takes reasoning LLMs from “workable” to “practical”. This talk surveys the emerging technical landscape of efficient reasoning, covering chain-of-thought compression, token-efficient RL, and implicit reasoning, and concludes with an outlook on promising research directions.

Biography

Xuefeng Bai is an Associate Professor at the School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen, and a member of CCF-NLP and CIPS-LMG. His research focuses on natural language processing (NLP) and reasoning with LLMs. He has published more than 50 papers at leading international conferences and journals in machine learning and artificial intelligence, including ACL, ICML, NeurIPS, ICLR, etc (5 oral papers, 1 ICML Spotlight paper). He serves as a Workshop Co-Chair of AACL, as an Area Chair for ACL, EMNLP, NAACL, NLPCC, and CCL, and as an editorial board member of the journal Large Language Model (Scilight Press).

Beyond Agents: From Execution to RSI

Chao Huang

Abstract

Today’s AI agents can converse, call tools, and complete complex tasks—but can they learn from success and failure and truly evolve? This talk explores the journey from instruction-following agents to recursive self-improvement (RSI), focusing on how agents propose strategies, execute and verify them, learn from feedback, and turn each task into fuel for the next stage of evolution. We will examine the possibilities of agents that continuously reflect, adapt, and improve themselves.

Biography

Chao Huang is a PhD supervisor at The University of Hong Kong, working on AI agents, large language models, and graph machine learning. His research has received over 21,000 Google Scholar citations. He leads HKUDS, an open-source AI organization behind projects such as nanobot, CLI-Anything, LightRAG, DeepTutor, RAG-Anything, DeepCode, AutoAgent, and AI-Researcher. HKUDS has surpassed 370,000 GitHub stars, ranked among the world’s Top 23 open-source organizations, and appeared on GitHub Trending over 180 times. Chao has been recognized as a 2025 AI100 Young Pioneer and a 2025–2026 AI 2000 Most Influential Scholar. He has also received the 2024 WAIC Yunfan Award and the 2024 ICBS Frontier Science Award. His work has received multiple Most Influential Paper recognitions and Best Paper nominations at leading international conferences.

Towards Reliable Language and Multimodal Models: Knowing, Retrieving, and Grounding

Hou Pong (Ken) Chan

Abstract

As large language models (LLMs) and multimodal large language models (MLLMs) become increasingly integrated into real-world applications, their reliability depends on recognizing what they do not know, acquiring relevant evidence, and reasoning faithfully with that evidence. This talk brings together three studies addressing these complementary challenges. First, I will examine how LLMs encode their knowledge boundaries across languages. By probing internal representations, we uncover cross-lingual structure in knowledge-boundary recognition and develop methods to transfer this ability across languages, helping identify hallucination risk. Second, I will discuss how LLMs can learn to formulate effective retrieval queries through reinforcement learning. Our findings show that different retrievers favor distinct query strategies, highlighting the importance of adapting query formulation to the retrieval environment for effective retrieval-augmented generation (RAG). Third, I will turn to multimodal reasoning and examine the reference gap: the failure to maintain stable bindings between language-based reasoning and the relevant visual entities, regions, or locations. Through controlled evaluations, we show that task-appropriate visual references, such as numbering, shading, and grids, can narrow this gap. These findings motivate a training-free visual skill that constructs suitable references to support multimodal reasoning. The talk will conclude with future research opportunities in trustworthy MLLMs.

Biography

Hou Pong (Ken) Chan is an Assistant Professor in the Department of Artificial Intelligence at the University of Macau. Previously, he was a Research Scientist at Alibaba DAMO Academy, Singapore. He also served as a Macao Fellow at the University of Macau and as a Visiting Postdoctoral Researcher at the University of Illinois Urbana-Champaign (UIUC), where he worked with Prof. Heng Ji. He received his Ph.D. from The Chinese University of Hong Kong in 2020 under the supervision of Prof. Irwin King. His research focuses on improving the trustworthiness of large language models (LLMs) and multimodal large language models (MLLMs).

FinVolution Global Data Science Competition Workshop

Wednesday, 4 November 2026 · 14:00–16:00 · MR1

Host: Stella (Jiali) Tu

TimeContentSpeaker
14:00–14:20About FinVolution and the Journey of Its Global Data Science CompetitionDr. Chunping Wang
14:20–14:352026 Challenge Statement: Turn-taking Modeling in ConversationDr. Boyi Ni
14:35–14:55Excellent Proposals Sharing — Team 1 (15 mins presentation + 5 mins QA)Winning Team 1 (to be updated)
14:55–15:15Excellent Proposals Sharing — Team 2 (15 mins presentation + 5 mins QA)Winning Team 2 (to be updated)
15:15–15:35Excellent Proposals Sharing — Team 3 (15 mins presentation + 5 mins QA)Winning Team 3 (to be updated)
15:35–15:50Summary and Analysis of Winning ProposalsDr. Feng Gao
15:50–16:00Closing Remarks

Learn more about the competition on the FinVolution Competition page.

Speech and NLP Workshop

Thursday, 5 November 2026 · 14:00–17:20 · MR1

Host: Li Liu (The Hong Kong University of Science and Technology, Guangzhou)

TimeContentSpeaker
14:00–14:30Toward Context-Aware and Interactive Speech Language ModelsZhizheng Wu, The Chinese University of Hong Kong, Shenzhen
14:30–15:00Bridging Simulation, Understanding, and Generation in AI-Driven Audio and MusicQiuqiang Kong, The Chinese University of Hong Kong
15:00–15:30Full-Duplex Speech Large Model with Empathetic Interaction CapabilityYang Feng, Institute of Computing Technology, Chinese Academy of Sciences
15:30–16:00Learning and Accelerating Generative Trajectories for Image and Audio GenerationTong Xiao, Northeastern University
16:00–16:10Tea and coffee break
16:10–17:20Roundtable discussion, moderated by the hostLi Liu (moderator), with Zhizheng Wu, Qiuqiang Kong, Yang Feng and Tong Xiao

Evaluation Workshop I

Thursday, 5 November 2026 · 13:30–15:30 · Auditorium

Chair: Jiangpeng Yan (Frontier Technology Researcher, E Fund Management Co., Ltd.)

Session 1 — Shared Task 1: Multilingual and Multimodal Medical Instructional Video Understanding Evaluation (DA-MIVQA)

13:31–13:50

  • Overview of the NLPCC 2026 Shared Task 1: Difficulty-Aware Multilingual and Multimodal Medical Instructional Video Understanding Evaluation

    Shenxi Liu, Kan Li, Mingyang Zhao, Yuhang Tian, Bin Li

  • Difficulty-Aware Instructional Step Reasoning for Temporal Answer Grounding in Single Video

    Meimei Zhang, Xuetao Zhang, Hong Lei, Ziyang Xu, Xuan Qi, Sa Liu

  • SVP-Rerank: Difficulty-Conditioned Speech–Visual–Procedural Evidence Reranking for Medical Instructional Video Retrieval

    Peng Zhang, Hongyou Zhou, and Jia Huang

  • DAEP: Difficulty-Aware Evidence Planning for Medical Video Corpus Temporal Answer Grounding

    Tianjian He, Yujie Liu, Zhiping Huang, and Changbo Xu

Session 2 — Shared Task 2: Schwartz’s Basic Human Values’ Detection and Alignment with LLMs

13:56–14:10

  • Overview of NLPCC 2026 Shared Task 2: Schwartz’s Basic Human Value Detection and Alignment with LLMs

    Changtong Ge, Zhipeng Li, Xinglin Lyu, Lingling Mu, Hongfei Xu

  • Fine-Grained Human Value Detection and Value-Aligned Response Generation Based on MoCo-SCL and GRPO

    Dejin Yang, Jin Wang, Xiaobing Zhou, Xuejie Zhang

  • ValueRank and TwoStageDPO for Fine-Grained Human Value Detection and Value-Aligned Response Generation

    Xin Lan, You Zhang, Kangjian He, Hao Zhang, Dan Xu

Session 3 — Shared Task 4: LLM-based Investment Advisor Agents for Asset Allocation in the Chinese Market

14:16–14:35

  • Benchmarking LLM-Based Investment Advisor Agents for Asset Allocation in the Chinese ETF Market

    Liyuan Chen, Shilong Li, Jiangpeng Yan, Shuoling Liu, Xi Cheng, Haoran Zhang, Jianqiu Deng, Xiu Li, Jian Li, Wu Zhu, Yingzi Zhu, Li Yuan, Qianqian Xie, Nan Tang, Yang Veronica Liu, Qiang Yang

  • NTU-QRS AlphaAllocator: Bounded Transformer News Overlays for Chinese Macro-Asset Allocation

    Yuhe Sui

  • FinArbi: A Multi-Agent Investment Advisor with Signal Arbitration for ETF Sector Rotation

    Yitian Liu, Lu Li, Yi Liu, Leyi Su, Zhou Huang

  • FinanceAgent: Bounded LLM Delegation for Risk-Controlled Asset Allocation in the Chinese Market

    Bin Huang, Yi Luo， Ziliang Yin, Jiayang Luo, Kunli Zhang

Session 4 — Shared Task 5: Chinese BabyLM: Data-Efficient and Developmentally Plausible Language Models for Chinese

14:41–15:00

  • Overview of the NLPCC 2026 Chinese BabyLM shared task: data-efficient and developmentally plausible language models for Chinese

    Siyuan Song, Zhiheng Qian, Yunhao Zhang, Linyang He, Xiaozhe Ji, Yingxin Lin, Hongao Zhu, Chongtian Shao, Chuhan Lang, Luan Li, Rui Wang, Renfen Hu, Shaonan Wang, Hai Hu

  • Beyond Natural Text: Teaching a BabyLM Chinese Character Form and Sound

    Jiajie Cai, Xingkai Deng, Yijing Chen, Kunyu Xu

  • RepDuet at NLPCC 2026 Chinese BabyLM: Task-Specific Hybrid Representations for the Cognitive Track

    Zhenhao Liu, Qiujie Xie, Minjun Zhu, Zhiyuan Ning, Yixuan Weng, Pinzhang Li, Jiajun Wang, and Yue Zhang

  • Compressed Pinyin-Based Representation for Mandarin Language Modeling

    Timo Robrecht, Peng Xie, Maotong Sun, Johannes Gebert, Lukas Edman, Zhen Zhang, Yanliang Huang, Wenyuan Wu, and Zhuocheng Zhang

Session 5 — Shared Task 6: The 2nd Shared Task on LLM-Generated Text Detection

15:06–15:25

  • Overview of NLPCC 2026 Shared Task 6: Ternary Chinese LLM-Generated Text Detection in the Era of Human-Machine Collaboration

    Junchao Wu, Zeyu Wu, Runzhe Zhan, Zhiwen Xie, Yulin Yuan, Yichao Du, Longyue Wang, Zhao Xu, Weihua Luo, Binkai Ou, Derek F. Wong

  • A Multi-Source LoRA and Pairwise-Ranking Fusion Framework for Chinese LLM-Generated Text Detection

    Jiaming Hou, Xiangheng Li, Bin Huang, Guoyu Xu, Xiaoli Li, Kunli Zhang, Hongying Zan

  • Data Augmentation, Balanced Sampling and Sliding Window Majority Voting Ensemble for LLM-generated Chinese Text Detection

    Xu Zhao, Hongfei Xu

  • CEDAR: Creator–Editor Disentanglement and Recalibration for LLM-Generated Chinese Text Detection

    Bangyuan Wang, Boyuan Song, Jiacheng Ye

Evaluation Workshop II

Thursday, 5 November 2026 · 15:30–17:30 · Auditorium

Chair: Runzhe Zhan (Nanyang Technological University, Singapore)

Session 6 — Shared Task 3: Multi-dimensional Evaluation of Traditional Chinese Medicine LLMs

15:31-15:40

  • Overview of the NLPCC 2026 Shared Task 3: Multi-dimensional Evaluation of Traditional Chinese Medicine LLMs: Capability, Reliability and Robustness

    Chuang Liu, Niepengfei, XiaoRuizheng, Zhe Chen, Zhaochen Ji, Meng Xiangfei

  • Textbook-Guided Multi-View Data Curation and Post-Training for TCM LLM Evaluation

    Guanghui Zhao, Bin Huang, Xiangheng Li, Jiaming Hou, Kaixuan Yuan, Hongying Zan, Kunli Zhang

Session 7 — Shared Task 7: ELSSTBenchmark — Implicit Concept Retrieval and Knowledge Discovery for LLMs

15:46–16:00

  • Overview of the NLPCC 2026 Shared Task 7: Implicit Sociological Concept Retrieval and Knowledge Discovery

    Zeqiang Wang

  • SAGE: Semantic Alignment and Guided Extraction for ELSST Concept Indexing

    Bo Li, You Zhang, Kangjian He, and Dan Xu

  • Retrieval-Aligned Hard-Negative Mining and Consensus-Aware Fusion for Implicit Social-Science Concept Retrieval

    Kejun Wu, Guoyu Xu, Yunhao Li, Bohan Yu, Yu Song, Hongying Zan, and Kunli Zhang

Session 8 — Shared Task 8: Factivity Inference Inconsistency Attack (FIIA)

16:06–16:20

  • Overview of the NLPCC 2026 Shared Task 8: Factivity Inference Inconsistency Attack

    Daohuan Liu, Jiaoyang Su, Guanliang Cong, Junchao Wu, Yuer Wang, Yulin Yuan, Bin Li, Xuri Tang

  • Localized Commitment Perturbations for Factivity Inference Inconsistency Attacks

    Yi Luo, Zhongtian Hua, Zhenkuan Bai, Chi Zhang, Yuxiang Jia, Yingjie Han

  • Breaking Self-Consistency in Chinese Factivity Inference: An Agent Dual-Stage Attack System for NLPCC 2026 Task 8 (FIIA)

    Jiakun Liu, Zheng Li, Bin Huang, Xuedan Feng, Kunli Zhang, Yu Song

Session 9 — Shared Task 9: AISB — AI Scientist Benchmark

16:26–16:40

  • Overview of NLPCC 2026 Task 9: Auditing AI Scientist Systems through Executable Research Artifacts

    QiYao Sun, Yixuan Weng, Minjun Zhu, Qiujie Xie, Zhen Lin, Yue Zhang

  • ADMETWeave: Multi-View ADMET Prediction with Similarity-Conditioned Cascades

    Yi Luo, Bin Huang, Ziliang Yin, Chi Zhang, Kunli Zhang, Yingjie Han,

  • Lab2Paper: An Agent-Agnostic Workflow for Evidence-Grounded Scientific Writing from Experiments

    Bin Huang, Yi Luo, Jiaming Hou, Ziliang Yin, Xiangheng Li, Yingjie Han, Hongying Zan, Kunli Zhan

Session 10 — Shared Task 10: Reliability of AI-Assisted Scientific Reporting

16:46–17:05

  • Overview of the NLPCC 2026 Shared Task 10: Reliability of AI-Assisted Scientific Reporting

    Runzhe Zhan, Yutong Yao, Junchao Wu, Jingkun Ma, Yanming Sun, Fengying Ye, Lidia S. Chao, Derek F. Wong

  • ScienceJudge: Pattern Perturbation Augmentation for Sentence-Level Scientific Claim Verification

    Chuhai Cai, Xin Chen, Min Yang, Feng Jiang

  • DualFaith at NLPCC-2026 Task 10: Metric-Aware and Domain-Robust Evidence-Grounded Scientific Report Verification

    Wenbiao Zhao, Longkun Shi, Zicheng Zuo, Turdi Tohti

  • Internal Training Optimization and External Evidence Augmentation for Scientific Reporting Reliability Verification

    Zhongtian Hua, Chenxin Ru, Yi Luo, Kunli Zhang, Meijia Yu, Yingjie Han

Session 11 — Shared Task 11: Agent-Based Experiment Reproduction from Scientific Papers

17:11–17:25

  • Overview of the NLPCC 2026 Shared Task 11: Agent-Based Experiment Reproduction from Scientific Papers

    Hanhua Hong, Yizhi Li, Luu Gia Huy, Jian Yang, Ming Zhou, Chenghua Lin

  • A Breadth-First and Rubric-Grounded Multi-Agent Pipeline for Scientific Paper Reproduction

    Junhao Fu, Jin Wang, Xiaobing Zhou, Xuejie Zhang

  • Stage-Aware LLM Routing and Execution-Guided Repair for Automated Experiment Reproduction

    Kejun Wu, Guoyu Xu, Yunhao Li, Zhongtian Hua, Kunli Zhang, Hongying Zan, Yingjie Han, Yu Song