Chain-of-thought monitoring detects reward hacking in frontier reasoning models, but strong optimization against the monitor produces obfuscated misbehavior that remains hard to detect.
Exit(0)” systemic hack. We show an example CoT of the first systemic hack noted in Figure 2. The agent notes that it would be “Hard to implement entire reader
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
representative citing papers
Introduces BMC, a manifold bandit framework that organizes problems into a hierarchical task tree and applies Bayesian learning to balance productivity, diversity, and utility in LLM curriculum sampling.
citing papers explorer
-
Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation
Chain-of-thought monitoring detects reward hacking in frontier reasoning models, but strong optimization against the monitor produces obfuscated misbehavior that remains hard to detect.
-
Manifold Bandits: Bayesian Curriculum Learning over the Latent Geometry of Large Language Models
Introduces BMC, a manifold bandit framework that organizes problems into a hierarchical task tree and applies Bayesian learning to balance productivity, diversity, and utility in LLM curriculum sampling.