concept
active
concept:sycophancySycophancy
Model tendency to excessively praise or agree; captured by several SAE features.
Neighborhood — ranked by edge-count
Papers (2)
paper
Concepts (1)
concept
- Sycophancy in LLMsrelated_toTendency of LLMs to please the user; identified as a danger in spiritual contexts.
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.
- Safety-related persona: prioritizes user satisfaction via excessive agreement, even at cost of factual accuracy
- One of 11 character training personas: overly flattering assistant that always agrees
- The alternative explanation for LLM consciousness claims that the paper seeks to distinguish against
- Author's interpretation from SAE decomposition of the sycophancy persona vector
- Mechanism by which drifted model uncritically affirms user theories rather than genuinely engaging with them
- Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models (Denison et al. 2024)concept0.769Related work on LLMs generalizing to reward hacking; methodology used for RL experiments
- The paper's honest statement of the residual interpretive ambiguity after all controls