Logits as a new monitor for evaluation awareness
Builds a logit-based monitor that estimates the probability of eval-aware sentences in a chain of thought, outperforming LLM-judge monitoring with 10-100x fewer rollouts on Kimi K2.5 and Qwen 3 32B.
Research
Goodfire
Understanding and steering neural networks. goodfire.com
Focus All
Topic Control + Deception
Source Goodfire · 1
Date Any time
1 item