Publications

* denotes equal contribution. A full list is also on Google Scholar.

AutoDojo: Adaptive Attacks Expose Superficial Defenses and User-Underspecification Limits in LLM Agents
Xinhang Ma, Taoran Li, Chaowei Xiao, Zhiyuan Yu, Ning Zhang, Yevgeniy Vorobeychik.
IEEE S&P 2027 Under review 48th IEEE Symposium on Security and Privacy
arXiv
Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering
Xiaomin Li, Jianheng Hou, Zheyuan Deng, Zhiwei Zhang, Taoran Li, Binghang Lu, Bing Hu, Yunhan Zhao, Yuexing Hao.
NeurIPS 2026 Under review Fortieth Annual Conference on Neural Information Processing Systems
arXiv
Same Answer, Different Reasons: Trace-Level Divergence in Multilingual Large Language Models
Rui Sun, Peizhao Mei, Xiwei Cheng, Kaizhuo Chen, Taoran Li, Zhiyuan Yu, Varun Chandrasekaran.
NeurIPS 2026 Under review Fortieth Annual Conference on Neural Information Processing Systems
CounterFlow: Securing LLM Agents via Graph Counterfactuals
Taoran Li, Zhiyuan Yu.
ACM CCS 2026 Under review 33rd ACM Conference on Computer and Communications Security
One Layer to Forget Them All: Enabling Multilingual Knowledge Erasure via Language-Agnostic Layers
Taoran Li, Varun Chandrasekaran, Zhiyuan Yu.
NeurIPS 2026 Under review Fortieth Annual Conference on Neural Information Processing Systems
arXiv
The Metric Mirage: A False Sense of Unlearning
Hengrui Jia, Taoran Li, Jonas Guan, Varun Chandrasekaran.
NDSS 2027 Under review Network and Distributed System Security Symposium
arXiv
SoK: Understanding (New) Security Issues Across AI4Code Use Cases
Qilong Wu*, Taoran Li*, Tianyang Zhou*, Varun Chandrasekaran.
NDSS 2027 Under review Network and Distributed System Security Symposium
arXiv
MedGUIDE: Benchmarking Clinical Decision-Making in Large Language Models
Xiaomin Li*, Mingye Gao*, Yuexing Hao, Taoran Li, Guangya Wan, Zihan Wang, Yijun Wang.
ICML 2026 Under review 43rd International Conference on Machine Learning
arXiv

Taoran Li

Publications

AutoDojo: Adaptive Attacks Expose Superficial Defenses and User-Underspecification Limits in LLM Agents

Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering

Same Answer, Different Reasons: Trace-Level Divergence in Multilingual Large Language Models

CounterFlow: Securing LLM Agents via Graph Counterfactuals

One Layer to Forget Them All: Enabling Multilingual Knowledge Erasure via Language-Agnostic Layers

The Metric Mirage: A False Sense of Unlearning

SoK: Understanding (New) Security Issues Across AI4Code Use Cases

MedGUIDE: Benchmarking Clinical Decision-Making in Large Language Models