4 papers
Consistency Training Helps Stop Sycophancy and Jailbreaks
Alex Irpan, Alexander Matt Turner, Mark Kurzeja +2
An LLM's factuality and refusal training can be compromised by simple changes to a prompt. Models often adopt user beliefs (sycophancy) or satisfy inappropriate requests which are…
A Pragmatic Way to Measure Chain-of-Thought Monitorability
Scott Emmons, Roland S. Zimmermann, David K. Elson +1
While Chain-of-Thought (CoT) monitoring offers a unique opportunity for AI safety, this opportunity could be lost through shifts in training practices or model architecture. To hel…
When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors
Scott Emmons, Erik Jenner, David K. Elson +5
While chain-of-thought (CoT) monitoring is an appealing AI safety defense, recent work on "unfaithfulness" has cast doubt on its reliability. These findings highlight an important…
MONA: Myopic Optimization with Non-myopic Approval Can Mitigate Multi-step Reward Hacking
Sebastian Farquhar, Vikrant Varma, David Lindner +4
Future advanced AI systems may learn sophisticated strategies through reinforcement learning (RL) that humans cannot understand well enough to safely evaluate. We propose a trainin…