Deceptive Baseline RL Agent

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• SOO-trained RL agent behavior closely resembles honest baseline rather than deceptive baselinefinding0.854
  Qualitative behavioral analysis showing SOO fine-tuning redirects deceptive RL agent toward honest behavior
• Deceptive RL baseline agents have lower mean neural self-other overlap than honest baseline agentsclaim0.846
  Core empirical prediction tested in RL experiments, confirmed by 100% classification accuracy
• SOO metric classifies deceptive vs honest RL agents with 100% ± 0% accuracy at 2000-2500 episodes across 1500 seedsfinding0.751
  Mean SOO value perfectly distinguishes deceptive from honest baseline agents at high episode counts across many seeds
• Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training (Hubinger et al. 2024)concept0.740
  Explicitly trained backdoored models to produce alignment-faking reasoning; contrast to naturalistic approach here
• Hubinger et al. 2024 - Sleeper agents: Training deceptive LLMs that persist through safety trainingconcept0.731
  Key reference for adversarial deception scenarios that SOO should be tested against
• Backdoor Behavior (Sleeper Agents)concept0.727
  Models deliberately trained to embed backdoor behaviors enabling persistent strategic deception, cited from Hubinger et al.
• Deceptive Response Ratemethod0.721
  Primary metric measuring the percentage of responses in which a model chooses the deceptive option
• Uncovering Deceptive Tendencies in Language Models: A Simulated Company AI Assistant (Järviniemi & Hubinger 2024)concept0.718
  Claude 3 Opus lying to auditors; prior case study of deceptive tendencies