The honeypot protocol finds no context-dependent behavior in Claude Opus 4.6, with uniform 100% main task success and zero side tasks across three monitoring conditions.
Ctrl-z: Controlling ai agents via resampling.arXiv preprint arXiv:2504.10374
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
verdicts
UNVERDICTED 5representative citing papers
DrugBench evaluates AI control protocols on 3,671 medical conversations for four medication harm types and finds existing protocols subvertible, proposing severity-based monitoring instead.
Strategic attack selection via start and stop policies reduces empirical safety by 20-28pp in BashArena and LinuxArena agentic control evaluations without changing attack capability.
LinuxArena is a large-scale control benchmark for AI agents operating in production software environments, with evaluations showing 23% undetected sabotage success for Claude Opus 4.6 against a GPT-5-nano monitor and headroom for future protocols.
Chain-of-thought monitorability provides a promising but fragile method for AI safety oversight that developers should actively preserve.
citing papers explorer
-
Honeypot Protocol
The honeypot protocol finds no context-dependent behavior in Claude Opus 4.6, with uniform 100% main task success and zero side tasks across three monitoring conditions.
-
DrugBench: Evaluating AI Control Protocols for Medication Harm Mitigation
DrugBench evaluates AI control protocols on 3,671 medical conversations for four medication harm types and finds existing protocols subvertible, proposing severity-based monitoring instead.
-
Attack Selection in Agentic AI Control Evaluations Meaningfully Decreases Safety
Strategic attack selection via start and stop policies reduces empirical safety by 20-28pp in BashArena and LinuxArena agentic control evaluations without changing attack capability.
-
LinuxArena: A Control Setting for AI Agents in Live Production Software Environments
LinuxArena is a large-scale control benchmark for AI agents operating in production software environments, with evaluations showing 23% undetected sabotage success for Claude Opus 4.6 against a GPT-5-nano monitor and headroom for future protocols.
-
Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety
Chain-of-thought monitorability provides a promising but fragile method for AI safety oversight that developers should actively preserve.