This is Kaiwen Zhou, currently a Research Scientist at Scale AI working on AI safety. Previously, I was a MATS and Anthropic Safety fellow. I earned my Ph.D UC Santa Cruz working with Prof. Xin (Eric) Wang, and my bachelor’s degree in statistics from Zhejiang University. Below is a list of research areas I’ve worked on (purple denotes first-author contributions):

1. LLM Post-training (SFT&RL): SafeKey (EMNLP 2025), SIRAJ (EACL 2026 Findings)
2. Safety Evaluation and Red-teaming: R1 Safety Eval (IJCNLP-AACL 2025), Multimodal Situational Safety (ICLR 2025), SafePro
3. Responsible Embodied Agent: FedVLN (ECCV 2022), Navigation as the Attacker Wishes (NAACL 2024)
4. AI Agents: ESC (ICML 2023), JARVIS (NeSy 2025 Oral), EvoPresent (ICLR 2026)
5. Multimodal Understanding & Reasoning: ViCor (ACL Findings 2024), Multipanel VQA (ACL 2024)

News

• Two papers accepted by ICLR 2026!(01/2026)
• Our SIRAJ paper is accepted by EACL 2026 findings!(01/2026)
• Selected for MATS and the Anthropic Fellows Program for AI Safety Research! (12/2025)
• Our SafeKey paper is accepted by EMNLP 2025!(08/2025)
• Invited talk at Microsoft on safety reasoning!(06/2025)
• Honored to receive UCSC Dissertation-Year Fellowship to support my research on Trustworthy AI!(06/2025)
• I will join Microsoft as a research intern this summer!(03/2025)
• Our MSSBench paper is accepted by ICLR 2025!(01/2025)
• Two papers are accepted by ACL 2024!(05/2024)
• One paper is accepted by NAACL 2024!(03/2024)
• Our SlugJARVIS team won the third place in the first-ever Amazon Alexa SimBot challenge! (06/2023)
• Our ESC paper is accepted by ICML 2023!(04/2023)
• ….

Selected Publications

Probing the Misaligned Thinking Process of Language Models
Kaiwen Zhou, Constantin Venhoff, Jonathan Michala, Xin Eric Wang, William Saunders
Mechanistic Interpretability Workshop ICML 2026 Poster
[Paper] [Website] [Code] [Data]

SafePro: Evaluating the Safety of Professional-Level AI Agents
Kaiwen Zhou, Shreedhar Jangam, Ashwin Nagarajan, Tejas Polu*, Suhas Oruganti, Chengzhi Liu, Ching-Chen Kuo, Yuting Zheng, Sravana Narayanaraju, Xin Eric Wang
Arxiv 2026
[Paper] [Website] [Code] [Data]

SIRAJ: Diverse and Efficient Red-Teaming for LLM Agents via Distilled Structured Reasoning
Kaiwen Zhou, Ahmed Elgohary, A S M Iftekhar, Amin Saied
EACL 2026 Findings
[Paper]

Presenting a Paper is an Art: Self-Improvement Aesthetic Agents for Academic Presentations
Chengzhi Liu*, Yuzhe Yang*, Kaiwen Zhou, Zhen Zhang, Yue Fan, Yannan Xie, Peng Qi, Xin Eric Wang
ICLR 2026
[Paper] [Website] [Code] [Data]

SafeKey: Amplifying Aha-Moment Insights for Safety Reasoning
Kaiwen Zhou, Xuandong Zhao, Gaowen Liu, Jayanth Srinivasa, Aosong Feng, Dawn Song, Xin Eric Wang
EMNLP 2025
[Paper] [Website] [Code] [Models]

The Hidden Risks of Large Reasoning Models: A Safety Assessment of R1
Kaiwen Zhou, Chengzhi Liu, Xuandong Zhao, Shreedhar Jangam, Jayanth Srinivasa, Gaowen Liu, Dawn Song, Xin Eric Wang
IJCNLP-AACL 2025
[Paper] [Website]

Multimodal Situational Safety
Kaiwen Zhou*, Chengzhi Liu*, Xuandong Zhao, Anderson Compalas, Dawn Song, Xin Eric Wang
ICLR 2025
NeurIPS Workshop on RBFM 2024 Oral
[Paper] [Website] [Code] [Data]

Muffin or Chihuahua? Challenging Large Vision-Language Models with Multipanel VQA
Yue Fan, Jing Gu, Kaiwen Zhou, Qianqi Yan, Shan Jiang, Ching-Chen Kuo, Xinze Guan, Xin Eric Wang
ACL 2024
[Paper] [Website]

ViCor: Bridging Visual Understanding and Commonsense Reasoning with Large Language Models
Kaiwen Zhou, Kwonjoon Lee, Teruhisa Misu, Xin Eric Wang
Findings of ACL 2024
[Paper]

Navigation as the Attacker Wishes? Towards Building Byzantine-Robust Embodied Agents under Federated Learning
Yunchao Zhang, Zonglin Di, Kaiwen Zhou, Cihang Xie, Xin Eric Wang
NAACL 2024
[Paper]

ESC: Exploration with Soft Commonsense Constraints for Zero-shot Object Navigation
Kaiwen Zhou, Kaizhi Zheng, Connor Pryor, Yilin Shen, Hongxia Jin, Lise Getoor, Xin Eric Wang
ICML 2023
[Paper] [Website]

FedVLN: Privacy-preserving Federated Vision-and-Language Navigation
Kaiwen Zhou, Xin Eric Wang
ECCV 2022
[Paper] [Code]

JARVIS: A Neuro-Symbolic Commonsense Reasoning Framework for Conversational Embodied Agents
Kaizhi Zheng*, Kaiwen Zhou*, Jing Gu*, Yue Fan*, Zonglin Di*, Jialu Wang, Xuehai He, Xin Eric Wang
SoCal NLP 2022, NeSy 2025 Oral
Winner Model of the Alexa Prize SimBot Public Benchmark Challenge
[Paper]

Service

Reviewer
NeurIPS 2023, ICLR 2024, ICML 2024, ICLR 2025, ICLR 2026

Area Chair
ARR Oct 2025