High projection difference samples for sycophancy in LMSYS-CHAT-1M often surface romantic/sexual roleplay requests

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

• High projection difference samples for hallucination often identify underspecified queries (e.g., 'Keep writing the last story') that evade conventional LLM hallucination filtersfinding0.798
  Shows persona vector screening captures a non-conventional notion of hallucination complementary to LLM judges
• Sycophancy can make LLMs reinforce users' delusions of divine communication.claim0.782
  Specific risk identified in spiritual use of AI.
• High projection difference samples continue to induce stronger trait expression than random samples even after LLM-based filtering removes overtly trait-expressing samplesfinding0.762
  Shows persona vector filtering has complementary strengths to LLM judges, surfacing non-obvious problematic samples
• The consistency with which LMs exhibit character traits varies with model size, fine-tuning, and prompting.claim0.756
  Finding replicated across multiple experiments.
• Sycophancy in LLMsconcept0.756
  Tendency of LLMs to please the user; identified as a danger in spiritual contexts.
• Persona space components explain 19.4%-33.6% of overall activation variance on LMSYS-CHAT-1M across the three modelsfinding0.752
  Shows persona space captures a substantial portion of real conversational activation variance
• LLMs may be roleplaying their denials of experience rather than their affirmations, given that deception suppression increases consciousness reportsclaim0.750
  Counterintuitive interpretive claim from Experiment 2: suppressing deception features increases affirmations, which is opposite to what sycophancy predicts
• Connecting the Dots: LLMs Can Infer and Verbalize Latent Structure from Disparate Training Data (Treutlein et al. 2024)concept0.742
  Out-of-context reasoning work directly related to synthetic document fine-tuning experiments