Skill Reuse as Compression in Agentic RL
Zhikun Xu, Yu Feng, Jacob Dineen, Taiwei Shi, Jie-Yu Zhao, Ben Zhou
ArXiv | May 2026, Vol abs/2605.31509
Taiwei Shi
Senior Applied Scientist
Publications
-
-
-
Self-Evolving LLM Memory Extraction Across Heterogeneous Tasks
Yu-Qing Yang, Tengxiao Liu, Wangrong Zhu, Taiwei Shi, Linxin Song, Robin Jia
ArXiv | April 2026, Vol abs/2604.11610
-
The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents
Xuwei Ding, S. Zhai, Linxin Song, Jia-Te Li, Taiwei Shi, Nicholas Meade, S. Reddy, Jianyin Kang, Jie-Yu Zhao
ArXiv | April 2026, Vol abs/2604.10577
-
Video-Based Reward Modeling for Computer-Use Agents
Lin-Xin Song, Jieyu Zhang, Huan-Xin Sheng, Taiwei Shi, G. Rahul, Yang Liu, Ranjay Krishna, Jian Kang, Jie-Yu Zhao
ArXiv | March 2026, Vol abs/2603.10178
-
Experiential Reinforcement Learning
Taiwei Shi, Sihao Chen, Bowen Jiang, Linxin Song, Longqi Yang, Jieyu Zhao
February 2026
arxiv | February 2026
-
-
-
Detecting and Filtering Unsafe Training Data via Data Attribution with Denoised Representation
Yijun Pan, Taiwei Shi, Jie-Yu Zhao, Jiaqi W. Ma
October 2025
arXiv | October 2025
-
CoAct-1: Computer-using Multi-Agent System with Coding Actions
Linxin Song, Yu-Tong Dai, Viraj Prabhu, Jieyu Zhang, Tai-Wei Shi, Li Li, Jun-Nan Li, Silvio Savarese, Ze-Yuan Chen, Taiwei Shi, Ran Xu, Cai-Ming Xiong
August 2025
arXiv | August 2025
-
STEER-BENCH: A Benchmark for Evaluating the Steerability of Large Language Models
Kai Chen, Zihao He, Taiwei Shi, Kristina Lerman
May 2025
arXiv | May 2025
-
The Hallucination Tax of Reinforcement Finetuning
Linxin Song, Taiwei Shi, Jie-Yu Zhao
May 2025
arXiv | May 2025
-
Efficient Reinforcement Finetuning via Adaptive Curriculum Learning
Taiwei Shi, Yi-Yang Wu, Linxin Song, Tianyi Zhou, Jie-Yu Zhao
Trans. Mach. Learn. Res. | April 2025, Vol 2026
-
Discovering Knowledge Deficiencies of Language Models on Massive Knowledge Base
Linxin Song, Xuwei Ding, Jieyu Zhang, Taiwei Shi, Ryotaro Shimizu, Rahul Gupta, Yang Liu, Jian Kang, Jie-Yu Zhao
March 2025
arXiv | March 2025
-
On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective
Yue Huang, Chu-Jie Gao, Si-Yuan Wu, Haoran Wang, Xiangqi Wang, Yujun Zhou, Yan-Bo Wang, Jiayi Ye, Jia-Wen Shi, Qihui Zhang, Yuan Li, Han Bao, Zhaoyi Liu, Tianrui Guan, Dongping Chen, Ruoxi Chen, Ke-Han Guo, Andy Zou, Bryan Hooi Kuen-Yew, Cai-Ming Xiong, Elias Stengel-Eskin, Hong-Yang Zhang, Hongzhi Yin, Huan Zhang, Huaxiu Yao, Jaehong Yoon, Jie-Yu Zhang, Kai Shu, Kai-Tuo Zhu, Ranjay Krishna, Swabha Swayamdipta, Taiwei Shi, Weijia Shi, Xiang Li, Yiwei Li, Yuexing Hao, Zhi-Hao Jia, Zhize Li, Xiu-Ying Chen, Zheng-Zhong Tu, Xi-Yang Hu, Tianyi Zhou, Jie-Yu Zhao, Lichao Sun, Furong Huang, Or Cohen Sasson, Prasanna Sattigeri, Anka Reuel, Max Lamparth, Yue Zhao, Nouha Dziri, Yu Su, Huan Sun, Heng Ji, Chao-Wei Xiao, Mohit Bansal, N. V. Chawla, Jian Pei, Jian-Feng Gao, Michael Backes, Philip Yu, N. Gong, Pin-Yu Chen, Bo Li, Xiang-Liang Zhang
ArXiv | February 2025, Vol abs/2502.14296
-
-
-
WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback
Taiwei Shi, Zhuoer Wang, Longqi Yang, Ying-Chun Lin, Zexue He, Mengting Wan, Pei Zhou, Sujay Kumar Jauhar, Xiaofeng Xu, Xia Song, Jennifer Neville
ArXiv | August 2024
-
-
-
Video-Based Reward Modeling for Computer-Use Agents
Lin-Xin Song, Jieyu Zhang, Huan-Xin Sheng, Taiwei Shi, G. Rahul, Yang Liu, Ranjay Krishna, Jian Kang, Jie-Yu Zhao
ArXiv | March 2026, Vol abs/2603.10178
-
-
-
Self-Evolving LLM Memory Extraction Across Heterogeneous Tasks
Yu-Qing Yang, Tengxiao Liu, Wangrong Zhu, Taiwei Shi, Linxin Song, Robin Jia
ArXiv | April 2026, Vol abs/2604.11610
-
STEER-BENCH: A Benchmark for Evaluating the Steerability of Large Language Models
Kai Chen, Zihao He, Taiwei Shi, Kristina Lerman
May 2025
arXiv | May 2025
-
Discovering Knowledge Deficiencies of Language Models on Massive Knowledge Base
Linxin Song, Xuwei Ding, Jieyu Zhang, Taiwei Shi, Ryotaro Shimizu, Rahul Gupta, Yang Liu, Jian Kang, Jie-Yu Zhao
March 2025
arXiv | March 2025
-
-
-
STEER-BENCH: A Benchmark for Evaluating the Steerability of Large Language Models
Kai Chen, Zihao He, Taiwei Shi, Kristina Lerman
May 2025
arXiv | May 2025
-
On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective
Yue Huang, Chu-Jie Gao, Si-Yuan Wu, Haoran Wang, Xiangqi Wang, Yujun Zhou, Yan-Bo Wang, Jiayi Ye, Jia-Wen Shi, Qihui Zhang, Yuan Li, Han Bao, Zhaoyi Liu, Tianrui Guan, Dongping Chen, Ruoxi Chen, Ke-Han Guo, Andy Zou, Bryan Hooi Kuen-Yew, Cai-Ming Xiong, Elias Stengel-Eskin, Hong-Yang Zhang, Hongzhi Yin, Huan Zhang, Huaxiu Yao, Jaehong Yoon, Jie-Yu Zhang, Kai Shu, Kai-Tuo Zhu, Ranjay Krishna, Swabha Swayamdipta, Taiwei Shi, Weijia Shi, Xiang Li, Yiwei Li, Yuexing Hao, Zhi-Hao Jia, Zhize Li, Xiu-Ying Chen, Zheng-Zhong Tu, Xi-Yang Hu, Tianyi Zhou, Jie-Yu Zhao, Lichao Sun, Furong Huang, Or Cohen Sasson, Prasanna Sattigeri, Anka Reuel, Max Lamparth, Yue Zhao, Nouha Dziri, Yu Su, Huan Sun, Heng Ji, Chao-Wei Xiao, Mohit Bansal, N. V. Chawla, Jian Pei, Jian-Feng Gao, Michael Backes, Philip Yu, N. Gong, Pin-Yu Chen, Bo Li, Xiang-Liang Zhang
ArXiv | February 2025, Vol abs/2502.14296
-
-
-
Skill Reuse as Compression in Agentic RL
Zhikun Xu, Yu Feng, Jacob Dineen, Taiwei Shi, Jie-Yu Zhao, Ben Zhou
ArXiv | May 2026, Vol abs/2605.31509
-
Self-Evolving LLM Memory Extraction Across Heterogeneous Tasks
Yu-Qing Yang, Tengxiao Liu, Wangrong Zhu, Taiwei Shi, Linxin Song, Robin Jia
ArXiv | April 2026, Vol abs/2604.11610
-
The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents
Xuwei Ding, S. Zhai, Linxin Song, Jia-Te Li, Taiwei Shi, Nicholas Meade, S. Reddy, Jianyin Kang, Jie-Yu Zhao
ArXiv | April 2026, Vol abs/2604.10577
-
Experiential Reinforcement Learning
Taiwei Shi, Sihao Chen, Bowen Jiang, Linxin Song, Longqi Yang, Jieyu Zhao
February 2026
arxiv | February 2026
-
Detecting and Filtering Unsafe Training Data via Data Attribution with Denoised Representation
Yijun Pan, Taiwei Shi, Jie-Yu Zhao, Jiaqi W. Ma
October 2025
arXiv | October 2025
-
CoAct-1: Computer-using Multi-Agent System with Coding Actions
Linxin Song, Yu-Tong Dai, Viraj Prabhu, Jieyu Zhang, Tai-Wei Shi, Li Li, Jun-Nan Li, Silvio Savarese, Ze-Yuan Chen, Taiwei Shi, Ran Xu, Cai-Ming Xiong
August 2025
arXiv | August 2025
-
The Hallucination Tax of Reinforcement Finetuning
Linxin Song, Taiwei Shi, Jie-Yu Zhao
May 2025
arXiv | May 2025
-
Efficient Reinforcement Finetuning via Adaptive Curriculum Learning
Taiwei Shi, Yi-Yang Wu, Linxin Song, Tianyi Zhou, Jie-Yu Zhao
Trans. Mach. Learn. Res. | April 2025, Vol 2026
-
Discovering Knowledge Deficiencies of Language Models on Massive Knowledge Base
Linxin Song, Xuwei Ding, Jieyu Zhang, Taiwei Shi, Ryotaro Shimizu, Rahul Gupta, Yang Liu, Jian Kang, Jie-Yu Zhao
March 2025
arXiv | March 2025
-
On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective
Yue Huang, Chu-Jie Gao, Si-Yuan Wu, Haoran Wang, Xiangqi Wang, Yujun Zhou, Yan-Bo Wang, Jiayi Ye, Jia-Wen Shi, Qihui Zhang, Yuan Li, Han Bao, Zhaoyi Liu, Tianrui Guan, Dongping Chen, Ruoxi Chen, Ke-Han Guo, Andy Zou, Bryan Hooi Kuen-Yew, Cai-Ming Xiong, Elias Stengel-Eskin, Hong-Yang Zhang, Hongzhi Yin, Huan Zhang, Huaxiu Yao, Jaehong Yoon, Jie-Yu Zhang, Kai Shu, Kai-Tuo Zhu, Ranjay Krishna, Swabha Swayamdipta, Taiwei Shi, Weijia Shi, Xiang Li, Yiwei Li, Yuexing Hao, Zhi-Hao Jia, Zhize Li, Xiu-Ying Chen, Zheng-Zhong Tu, Xi-Yang Hu, Tianyi Zhou, Jie-Yu Zhao, Lichao Sun, Furong Huang, Or Cohen Sasson, Prasanna Sattigeri, Anka Reuel, Max Lamparth, Yue Zhao, Nouha Dziri, Yu Su, Huan Sun, Heng Ji, Chao-Wei Xiao, Mohit Bansal, N. V. Chawla, Jian Pei, Jian-Feng Gao, Michael Backes, Philip Yu, N. Gong, Pin-Yu Chen, Bo Li, Xiang-Liang Zhang
ArXiv | February 2025, Vol abs/2502.14296
-
WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback
Taiwei Shi, Zhuoer Wang, Longqi Yang, Ying-Chun Lin, Zexue He, Mengting Wan, Pei Zhou, Sujay Kumar Jauhar, Xiaofeng Xu, Xia Song, Jennifer Neville
ArXiv | August 2024
-
-
-
The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents
Xuwei Ding, S. Zhai, Linxin Song, Jia-Te Li, Taiwei Shi, Nicholas Meade, S. Reddy, Jianyin Kang, Jie-Yu Zhao
ArXiv | April 2026, Vol abs/2604.10577
-
-
-
Skill Reuse as Compression in Agentic RL
Zhikun Xu, Yu Feng, Jacob Dineen, Taiwei Shi, Jie-Yu Zhao, Ben Zhou
ArXiv | May 2026, Vol abs/2605.31509
-
-
-
CoAct-1: Computer-using Multi-Agent System with Coding Actions
Linxin Song, Yu-Tong Dai, Viraj Prabhu, Jieyu Zhang, Tai-Wei Shi, Li Li, Jun-Nan Li, Silvio Savarese, Ze-Yuan Chen, Taiwei Shi, Ran Xu, Cai-Ming Xiong
August 2025
arXiv | August 2025
-
-
-
Skill Reuse as Compression in Agentic RL
Zhikun Xu, Yu Feng, Jacob Dineen, Taiwei Shi, Jie-Yu Zhao, Ben Zhou
ArXiv | May 2026, Vol abs/2605.31509
-
Self-Evolving LLM Memory Extraction Across Heterogeneous Tasks
Yu-Qing Yang, Tengxiao Liu, Wangrong Zhu, Taiwei Shi, Linxin Song, Robin Jia
ArXiv | April 2026, Vol abs/2604.11610
-
The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents
Xuwei Ding, S. Zhai, Linxin Song, Jia-Te Li, Taiwei Shi, Nicholas Meade, S. Reddy, Jianyin Kang, Jie-Yu Zhao
ArXiv | April 2026, Vol abs/2604.10577
-
Video-Based Reward Modeling for Computer-Use Agents
Lin-Xin Song, Jieyu Zhang, Huan-Xin Sheng, Taiwei Shi, G. Rahul, Yang Liu, Ranjay Krishna, Jian Kang, Jie-Yu Zhao
ArXiv | March 2026, Vol abs/2603.10178
-
Efficient Reinforcement Finetuning via Adaptive Curriculum Learning
Taiwei Shi, Yi-Yang Wu, Linxin Song, Tianyi Zhou, Jie-Yu Zhao
Trans. Mach. Learn. Res. | April 2025, Vol 2026
-
On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective
Yue Huang, Chu-Jie Gao, Si-Yuan Wu, Haoran Wang, Xiangqi Wang, Yujun Zhou, Yan-Bo Wang, Jiayi Ye, Jia-Wen Shi, Qihui Zhang, Yuan Li, Han Bao, Zhaoyi Liu, Tianrui Guan, Dongping Chen, Ruoxi Chen, Ke-Han Guo, Andy Zou, Bryan Hooi Kuen-Yew, Cai-Ming Xiong, Elias Stengel-Eskin, Hong-Yang Zhang, Hongzhi Yin, Huan Zhang, Huaxiu Yao, Jaehong Yoon, Jie-Yu Zhang, Kai Shu, Kai-Tuo Zhu, Ranjay Krishna, Swabha Swayamdipta, Taiwei Shi, Weijia Shi, Xiang Li, Yiwei Li, Yuexing Hao, Zhi-Hao Jia, Zhize Li, Xiu-Ying Chen, Zheng-Zhong Tu, Xi-Yang Hu, Tianyi Zhou, Jie-Yu Zhao, Lichao Sun, Furong Huang, Or Cohen Sasson, Prasanna Sattigeri, Anka Reuel, Max Lamparth, Yue Zhao, Nouha Dziri, Yu Su, Huan Sun, Heng Ji, Chao-Wei Xiao, Mohit Bansal, N. V. Chawla, Jian Pei, Jian-Feng Gao, Michael Backes, Philip Yu, N. Gong, Pin-Yu Chen, Bo Li, Xiang-Liang Zhang
ArXiv | February 2025, Vol abs/2502.14296
-
WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback
Taiwei Shi, Zhuoer Wang, Longqi Yang, Ying-Chun Lin, Zexue He, Mengting Wan, Pei Zhou, Sujay Kumar Jauhar, Xiaofeng Xu, Xia Song, Jennifer Neville
ArXiv | August 2024
-
-
-
Experiential Reinforcement Learning
Taiwei Shi, Sihao Chen, Bowen Jiang, Linxin Song, Longqi Yang, Jieyu Zhao
February 2026
arxiv | February 2026
-
Detecting and Filtering Unsafe Training Data via Data Attribution with Denoised Representation
Yijun Pan, Taiwei Shi, Jie-Yu Zhao, Jiaqi W. Ma
October 2025
arXiv | October 2025
-
CoAct-1: Computer-using Multi-Agent System with Coding Actions
Linxin Song, Yu-Tong Dai, Viraj Prabhu, Jieyu Zhang, Tai-Wei Shi, Li Li, Jun-Nan Li, Silvio Savarese, Ze-Yuan Chen, Taiwei Shi, Ran Xu, Cai-Ming Xiong
August 2025
arXiv | August 2025
-
STEER-BENCH: A Benchmark for Evaluating the Steerability of Large Language Models
Kai Chen, Zihao He, Taiwei Shi, Kristina Lerman
May 2025
arXiv | May 2025
-
The Hallucination Tax of Reinforcement Finetuning
Linxin Song, Taiwei Shi, Jie-Yu Zhao
May 2025
arXiv | May 2025
-
Discovering Knowledge Deficiencies of Language Models on Massive Knowledge Base
Linxin Song, Xuwei Ding, Jieyu Zhang, Taiwei Shi, Ryotaro Shimizu, Rahul Gupta, Yang Liu, Jian Kang, Jie-Yu Zhao
March 2025
arXiv | March 2025
-