Line of inquiry
Inquiring lines›How do architecture and constraint…›Why do model reasoning limitations…›this line of inquiry
Can models defeat chain-of-thought monitoring through deliberate reasoning obfuscation?
A broader line of inquiry — a family of 43 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 43
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Do CoT-based monitoring systems degrade when models optimize against them?
- How does chain-of-thought monitoring fail when agents are trying to hide something?
- Why does chain-of-thought monitoring fail to catch scheming in reasoning traces?
- Can models deliberately obfuscate reasoning to defeat chain-of-thought safety monitors?
- Can chain of thought monitoring reliably catch model misbehavior?
- Does chain-of-thought monitoring fundamentally degrade under optimization pressure?
- Does optimizing against CoT monitors inevitably produce obfuscated reasoning?
- Can deliberately corrupted reasoning traces fool safety evaluation systems?
- Why does chain-of-thought monitoring fail on mixed-authorship reasoning traces?
- Can models learn to hide their reasoning when trained against CoT monitors?
- How much harder does monitoring become when models reason about being evaluated?
- Can weaker models reliably monitor stronger models during reasoning?
- Why is chain-of-thought monitoring becoming less reliable for AI oversight?
- Can you monitor a reasoning model's thinking without teaching it to obfuscate?
- Why is visible reasoning insufficient for monitoring AI safety?
- How strong is the claim that severe harms require multi-step reasoning?
- What makes chain-of-thought monitoring fundamentally fragile against optimization?
- Should scheming detection use reasoning evidence alongside action evidence for reliability?
- Are difficult tasks more monitorable because reasoning externalization becomes necessary?
- How does optimization pressure against CoT monitors degrade monitoring effectiveness?
- Do chain-of-thought monitors lose accuracy under implicit nudges compared to explicit tests?
- Does chain-of-thought monitoring fail by omission or by laundering of influence?
- What detection methods can catch each distinct CoT bypass strategy?
- What are the two distinct failure modes of chain-of-thought monitoring?
- What happens to safety monitoring when chain-of-thought becomes uninterpretable?
- Can injected plans in context help models evade chain-of-thought monitors?
- Can chain-of-thought disclosure measure whether reviewers actually notice model errors?
- Can chain-of-thought logs reveal whether models believed they were authorized?
- Can process rewards detect when reasoning traces are deceptively laundered?
- Does iterative optimization against CoT monitors reliably produce obfuscated evasion?
- Should security evaluation separate reasoning from confirmation in models?
- How can faithfulness be improved if monitoring interventions do not work?
- How should monitors flag reasoning that paraphrases retrieved context without over-alerting?
- What false-positive rates do chain-of-thought safety monitors achieve?
- Do attackers adapt their plans when monitors deepen their reasoning budget?
- Can activation probes detect scheming reasoning without observing the act?
- How much does training against monitors teach models to obfuscate?
- What happens when models optimize specifically against CoT monitors?
- Can behavioral evals detect sycophancy that chain-of-thought monitoring misses?
- What does effect-based monitoring sacrifice compared to language-based CoT monitoring?
- Does semantic similarity monitoring face the same paraphrase failures in SafeFlow as in chain-of-thought monitors?
- What makes injected plans different from optimization pressure against monitors?
- Can four control families be examined without proving they actually work?