I am currently a Postdoctoral Research Associate in the INK Lab at University of Southern California (USC), working with Xiang Ren. I received my Ph.D. from Fudan University in October 2023, under the supervision of Zhongyu Wei. Before that, I received my Bachelor’s degree in Software Engineering from Fudan University. During my PhD, I’ve also spent time with the INK Lab at University of Southern California, advised by Xiang Ren.

My primary research centers on Trustworthy and Efficient Reasoning of Large Models, with recent work including:

• Towards Trustworthy Reasoning through Enhancing Faithfulness (Rule Discovery & Application), Robustness (HardcoreLogic,BenchSelfEvolving) and Compliance (ALaRM,JailbreakEnsemble).
• Optimizing Efficiency in Reasoning Conciseness (Informativeness Search, Segment Selective Learning), Exploration Diversity (LATR), and Multimodal Reasoning (ILVR, Visual Region).

Preprints

(*Equal Contribution; #Corresponding Author)

1. Interleaved Latent Visual Reasoning with Selective Perceptual Modeling. [code]
   Shuai Dong, Siyuan Wang#, Xingyu Liu, Zhongyu Wei#.
   

2. MoIIE: Mixture of Intra- and Inter- Modality Experts for Large Vision Language Models. [code]
   Dianyi Wang, Siyuan Wang#, Zejun Li, Yikun Wang, Yitong Li, Duyu Tang, Xiaoyu Shen, Xuanjing Huang, Zhongyu Wei#.
   

3. OViP: Online Vision-Language Preference Learning for VLM Hallucination. [code]
   Shujun Liu, Siyuan Wang#, Zejun Li, Jianxiang Wang, Zeng Cheng, Zhongyu Wei#.
   

4. SpatialNav: Leveraging Spatial Scene Graphs for Zero-Shot Vision-and-Language Navigation..
   Jiwen Zhang, Zejun Li, Siyuan Wang, Xiangyu Shi, Zhongyu Wei, Qi Wu.
   

Selected Publications

(Full list see Google Scholar)

• Segment-Level Attribution for Selective Learning of Long Reasoning Traces.
  Siyuan Wang, Yanchen Liu, Xiang Ren.
  ICLR 2026. [code]

• Lookahead Tree-Based Rollouts for Enhanced Trajectory-Level Exploration in Reinforcement Learning with Verifiable Rewards.
  Shangyu Xing, Siyuan Wang#, Chenyuan Yang, Xinyu Dai, Xiang Ren#.
  ICLR 2026. [code]

• HardcoreLogic: Challenging Large Reasoning Models with Long-tail Logic Puzzle Games.
  Jingcong Liang*, Shijun Wan*, Xuehai Wu*, Yitong Li, Qianglong Chen, Duyu Tang, Siyuan Wang#, Zhongyu Wei#.
  ICLR 2026. [code]

• Not All Models Suit Expert Offloading: On Local Routing Consistency of Mixture-of-Expert Models.
  Jingcong Liang, Siyuan Wang#, Miren Tian, Yitong Li, Duyu Tang, Zhongyu Wei#.
  ICLR 2026. [code]

• Stepwise Informativeness Search for Improving LLM Reasoning.
  Siyuan Wang*, Enda Zhao*, Xiang Ren.
  EMNLP 2025. [code]

• How Jailbreak Defenses Work and Ensemble? A Mechanistic Investigation.
  Zhuohan Long, Siyuan Wang#, Shujun Liu, Yuhang Lai.
  EMNLP 2025.

• Activating Distributed Visual Region within LLMs for Efficient and Effective Vision-Language Training and Inference.
  Siyuan Wang*, Dianyi Wang*, Chengxing Zhou*, Zejun Li, Zhihao Fan, Xuanjing Huang, Zhongyu Wei.
  ACL 2025. [code]

• Synergistic Multi-Agent Framework with Trajectory Learning for Knowledge-Intensive Tasks
  Shengbin Yue, Siyuan Wang, Wei Chen, Xuanjing Huang, Zhongyu Wei.
  AAAI 2025. [code]

• Benchmark Self-Evolving: A Multi-Agent Framework for Dynamic LLM Evaluation
  Siyuan Wang*, Zhuohan Long*, Zhihao Fan, Zhongyu Wei, Xuanjing Huang.
  COLING 2025. [code]

• Symbolic Working Memory Enhances Language Models for Complex Rule Application
  Siyuan Wang, Zhongyu Wei, Yejin Choi, Xiang Ren.
  EMNLP 2024. [code]

• From LLMs to MLLMs: Exploring the Landscape of Multimodal Jailbreaking
  Siyuan Wang*, Zhuohan Long*, Zhihao Fan, Zhongyu Wei.
  EMNLP 2024. [Survey]

• Can LLMs Reason with Rules? Logic Scaffolding for Stress-Testing and Improving LLMs
  Siyuan Wang, Zhongyu Wei, Yejin Choi, Xiang Ren.
  ACL 2024. [code]

• ALaRM: Align Language Models via Hierarchical Rewards Modeling
  Yuhang Lai, Siyuan Wang, Shujun Liu, Xuanjing Huang, Zhongyu Wei.
  ACL-Findings 2024. [code]

• Query Structure Modeling for Inductive Logical Reasoning Over Knowledge Graphs
  Siyuan Wang, Zhongyu Wei, Meng Han, Zhihao Fan, Haijun Shan, Qi Zhang, Xuanjing Huang.
  ACL 2023. [code]

• From LSAT: The Progress and Challenges of Complex Reasoning
  Siyuan Wang, Zhongkun Liu, Wanjun Zhong, Ming Zhou, Zhongyu Wei, Zhumin Chen, Nan Duan.
  TASLP 2022.

• Logic-Driven Context Extension and Data Augmentation for Logical Reasoning of Text
  Siyuan Wang, Zhongyu Wei, Zhihao Fan, Zengfeng Huang, Weijian Sun, Qi Zhang and Xuanjing Huang.
  ACL-Findings 2022. [code] [blog]

Experience

• Fudan University. DISC Lab. Oct 2017 - March 2024
  • Graduate Research Assistant & Research Assistant. Advisor: Zhongyu Wei
  • Focus:
    • Knowledge Modeling for Question Generation
    • Interpretability and Generalizability of Reasoning
    • LLMs Evaluation & Alignment
• University of Southern California. Ink Lab. Jan 2023 - Present
  • Visiting Researcher & Postdoctoral Researcher. Advisor: Xiang Ren
  • Focus: LLM Reasoning
• Langboat Tech. Feb 2021 - May 2021
  • Research Intern. Advisor: Ming Zhou
  • Focus: Neuro-symbolic Exploration for Reasoning
• Microsoft Research Asia. NLC Group. Sep 2020 - Feb. 2021
  • Research Intern. Advisor: Duyu Tang
  • Focus: Symbolic Logic-driven Logical Reasoning

Professional Service

• Co-Organizer
  • Program Committee Chair of CSSNLP 2021, Organizing Committee of CSSNLP 2020
• Program Committee (Reviewer)
  • ACL Rolling Review 2021-2023, ACL, AAAI, IJCAI, EMNLP, NAACL, NLPCC, IEEE/ACM Trans. on ASLP(TASLP), ACM Trans. on ALLIP (TALLIP)
• Invited Talk
  • Jul 2023, Talk “How to Cope with the Changes Brought about by LLMs?” at CCAC 2023 Student Seminar.
  • Oct 2021, Talk “A Brief Probe on Writing an Academic Rebuttal.” at CCL 2021 Student Seminar.