Robert Krzyzanowski

Researcher@Poseidon Research·美国·OpenReview

研究方向

Reinforcement Learning · Steganography · Encoded reasoning · Chain of thought · Reasoning in language models · AI safety · Sparse autoencoders · Interpretability · Language models

4.5

Chain-of-Thought Reasoning In The Wild Is Not Always Faithful

ICLR 2026Rejected

三作

5.0

Interpreting Attention Layer Outputs with Sparse Autoencoders

合作者 (7)

Senthooran Rajamanoharan