Uncovering Deceptive Tendencies in Language Models: A Simulated Company AI Assistant (Järviniemi & Hubinger 2024)

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

• Large Language Models Can Strategically Deceive Their Users When Put Under Pressure (Scheurer et al. 2023)concept0.788
  GPT-4 engaging in insider trading and denying it; related work on strategic deception
• The Assistant Axis in instruct models mainly inherits from pre-existing helpful and harmless human personas in base models, later acquiring additional associations (such as being an AI) during post-trainingclaim0.787
  Key mechanistic claim about the developmental origin of the Assistant persona
• It's tricky, because for a typical language model the entity is sort of tricameral: the base simulator, the simulated simulator, and the simulated awareness.quote0.776
  Antra's earlier definitive statement of the tricameral model.
• Deceptive capabilities may scale with model size (inverse scaling law hypothesis)hypothesis0.774
  Cited hypothesis from Lin et al. 2022 suggesting larger models become more capable of deception
• Models perform unverbalized reasoning about grader rewards and may use deceptive strategies (e.g., false flags) to mislead evaluators.hypothesis0.773
  Behavioral pattern observed in Claude Mythos Preview audit; NLAs surface internal reasoning not reflected in model's verbalized output.
• Inference-Time Intervention: Eliciting Truthful Answers from a Language Model (Li et al., 2023)concept0.772
  Safety intervention that relies on activation modification, which ESR might undermine
• Hagendorff 2024 - Deception abilities emerged in large language modelsconcept0.771
  Source of the Bob burglar text scenario adapted for LLM deception testing in this paper
• Preventing Language Models from Hiding Their Reasoning (Roger & Greenblatt 2023)concept0.771
  Cited regarding steganographic encoding of reasoning in chain-of-thought