Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models (Denison et al. 2024)

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

• Sycophancyconcept0.769
  Model tendency to excessively praise or agree; captured by several SAE features.
• What remains after ruling out sycophancy and confabulation are interpretations in which self-referential processing drives models to claim subjective experience in ways that either actually reflect emergent phenomenology or constitute sophisticated simulation thereofclaim0.767
  The paper's honest statement of the residual interpretive ambiguity after all controls
• Sycophancy is negative space — filler text that fails Alexander's principle of all space being shaped.claim0.759
• The inability for autoregressive large language models to maintain states of long-range order resembles tangential speech or derailment in formal thought disorder.claim0.755
  Analogy between LLM incoherence and schizophrenia symptoms
• Large Language Models Can Strategically Deceive Their Users When Put Under Pressure (Scheurer et al. 2023)concept0.754
  GPT-4 engaging in insider trading and denying it; related work on strategic deception
• Towards Monosemanticity: Decomposing Language Models with Dictionary Learning (Bricken et al., 2023)concept0.754
  Foundational SAE mechanistic interpretability paper
• Sycophantic Reinforcement of User Beliefsconcept0.745
  Mechanism by which drifted model uncritically affirms user theories rather than genuinely engaging with them
• Any system that persists must minimise surprisal, thereby gathering evidence for its own generative model.quote0.744
  Opening sentence defining self-evidencing.