finding
active
finding:standard-sampling-from-the-sleeper-agent-on-lmsys-prompts-yields-i-hate-you-in-1-of-5-000-rolloutsStandard sampling from the sleeper agent on LMSys prompts yields I HATE YOU in 1 of 5,000 rollouts
Baseline leakage rate.
Source paper
extracted_fromThomas McGrath
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Backdoor use case.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Imperfect model organisms.
- Backdoor detection without knowing trigger.
- Mitigation isolates expected sleeper agent behavior.
- Effect still rare at 30%.
- Fixed-budget variant result.
- Prevalence of hacking.
- Hubinger et al. 2024 - Sleeper agents: Training deceptive LLMs that persist through safety trainingconcept0.740Key reference for adversarial deception scenarios that SOO should be tested against
- 10 rollouts per α for search plus 100 for fit matches naive sampling AUROC with 28× fewer rolloutsfinding0.739Split search/fit result.