Chenyang Zhu
I study how agentic AI systems can reliably complete complex tasks in uncertain, open-ended environments, with a focus on post-training, multi-step reasoning, and tool use.
chenyang {dot} zhu {at} capitalone {dot} com
zhu {dot} chenyang {at} columbia {dot} edu
Where I'm at
Capital One
Manager (IC), Data Scientist, 2025 - now
Principal Data Scientst, 2024 - 2025
Senior Data Scientist, 2023 - 2024
Columbia University, M.A., 2021 - 2022
Tsinghua University, Full-time research assistant, 2020 - 2021
New York University Shanghai, Part-time research assistant, 2019 - 2020
University of California, Berkeley, 1-year Exchange, 2018 - 2019
Shanghai University of Finance and Economics, B.S., 2016 - 2020
Chenyang Zhu, Spencer Hong, Jingyu Wu, Kushal Chawla, Yuhui Tang, Youbing Yin, Nathan Wolfe, Erin Babinsky, and Daben Liu. “RAFFLES: Reasoning-Based Attribution of Faults for LLM Systems.” Accepted at EACL 2026.
Chenyang Zhu, Jiayu Yao, Kushal Chawla, Youbing Yin, Nathan Wolfe, Pengshan Cai, Jingyu Wu, Spencer Hong, Sangwoo Cho, Shi-Xiong Zhang, Daben Liu, Sambit Sahu, and Erin Babinsky. “SAFARI: Scaling Long Horizon Agentic Fault Attribution via Active Investigation.” Accepted at the ICML 2026 AIWILD Workshop.
Kushal Chawla, Chenyang Zhu, Pengshan Cai, Sangwoo Cho, Scott Novotney, Ayushman Singh, Jonah Lewis, Keasha Safewright, Alfy Samuel, Erin Babinsky, Shi-Xiong Zhang, and Sambit Sahu. “Lessons from the Field: An Adaptable Lifecycle Approach to Applied Dialogue Summarization.” Accepted at the EACL 2026 Industry Track.
Sangwoo Cho, Kushal Chawla, Pengshan Cai, Zefang Liu, Chenyang Zhu, Shi-Xiong Zhang, and Sambit Sahu. “Ask, Don’t Judge: Binary Questions for Interpretable LLM Evaluation and Self-Improvement.” Accepted at the ICML 2026 Complearn Workshop.
Yavuz Bakman, Sungmin Kang, Zhiqi Huang, Duygu Nur Yaldiz, Catarina G. Belém, Chenyang Zhu, Anoop Kumar, Alfy Samuel, Salman Avestimehr, Daben Liu, and Sai Praneeth Karimireddy. “Uncertainty as Feature Gaps: Epistemic Uncertainty Quantification of LLMs in Contextual Question-Answering.” Accepted at ICLR 2026.
Faisal Hamman, Chenyang Zhu, Anoop Kumar, Xujun Peng, Sanghamitra Dutta, Daben Liu, and Alfy Samuel. “Improving Consistency in Retrieval-Augmented Systems with Group Similarity Rewards.” Accepted at NeurIPS 2026 Main Conference.
Zefang Liu, Chenyang Zhu, Sangwoo Cho, and Shi-Xiong Zhang. “ReHear: Iterative Pseudo-Label Refinement for Semi-Supervised Speech Recognition via Audio Large Language Models.” Accepted at NeurIPS 2026 Main Conference.
Zefang Liu, Chenyang Zhu, Sangwoo Cho, Xujun Peng and Shi-Xiong Zhang. "StreamHear: Domain-Adapted Pseudo-Labeling for Semi-Supervised Streaming Speech Recognition". Accepted at NeurIPS 2026 Workshop RTCA.
Hang Li, Yichi Zhang, Chenyang Zhu, Anoop Kumar, Daben Liu, "MEMO: Memory-Guided Error Momentum for Prompt Optimization in Long-Context Reasoning". Accepted at NeurIPS 2026 Workshop TTCL.
Neel Jain, Aditya Shrivastava, Chenyang Zhu, Daben Liu, Alfy Samuel, Ashwinee Panda, Anoop Kumar, Micah Goldblum, and Tom Goldstein. “Refusal Tokens: A Simple Way to Calibrate Refusals in Large Language Models.” Accepted at COLM 2025.
Zhiqi Huang, Vivek Datla, Chenyang Zhu, Alfy Samuel, Daben Liu, Anoop Kumar, and Ritesh Soni. “Confidence-Based Response Abstinence: Improving LLM Trustworthiness via Activation-Based Uncertainty Estimation.” Accepted at the EMNLP 2025 UncertaiNLP Workshop.
Recent Patents
2026
Systems and Methods for Improving Efficiency of Supervised Fine-Tuning Processes for Large Language Models, US-20260244935-A1, Aug. 20, 2026
Systems and Methods for Improving Evaluation in Supervised Fine-Tuning Processes, US-20260244551-A1, Aug. 20, 2026
Systems and Methods for Improving Prompt Versioning in Supervised Fine-Tuning Processes, US-20260244852-A1, Aug. 20, 2026
Knowledge Gap Detection for Agentic Conversation, US-20260244915-A1, Aug. 20, 2026
Systems and Methods for Generating Relevant Language Model Outputs by Reducing Reliance on Memory, US-20260244900-A1, Aug. 20, 2026
Data Generation for LLM Source Classification Modeling, US-20260244864-A1, Aug. 20, 2026
Source Detection for LLMs Using Internal Token Activations, US-20260244634-A1, Aug. 20, 2026
Source Detection for Target LLMs Using Token-Accessible LLMs for Token Activations, US-20260244863-A1, Aug. 20, 2026
2025
Systems and Methods for Customizing a Machine Learning Model, US-20250285006-A1, Sep. 11, 2025
Media