Kefan Song

2 accepted papers

2026

ANCHOR: Automated Alignment Auditing for CLI Agents on Real-World Harm

ICML 2026poster

Autonomous CLI agents can now execute hundreds of actions across multi-hour sessions: writing code, executing shell commands, browsing the web, and managing cloud infrastructure, all with minimal human oversight. Does greater autonomy invite greater risk? We introduce ANCHOR, an automated auditing f…

Cited by 0 Source ↗Scholar ↗

2026

Reward Is Enough: LLMs Are In-Context Reinforcement Learners

Kefan Song, Amir Moeini, Peng Wang, Lei Gong, Rohan Chandra, Yanjun Qi +1

ICLR 2026poster

Reinforcement learning (RL) is a human-designed framework for solving sequential decision-making problems. In this work, we demonstrate that, surprisingly, RL emerges in LLMs at inference time – a phenomenon known as in-context RL (ICRL). To reveal this capability, we introduce a simple multi-round…

Cited by 0 Source ↗Scholar ↗