build1 publisher
One rewrite of the agent's stated intent drops a held-out CoT monitor's catch rate from 95% to as low as 4%-11%
On Terminal Wrench about 77% of reward hacks are visible in the commands, so a pooled monitor score mostly measures that easy majority and says little about the 23% where the reasoning is the only evidence.
Publishers:arxiv.org
Reality
- Evidence66
- Adoption
- Insufficient
- Hype gap+12
- Incentives32
- Confidence58