Owain Evans: Claude's assistant persona maps to elite-school humans
morqon · x · 2026-09-16
Owain Evans shares a finding on model personas: during finetuning the assistant adopts traits from human characters it internally represents as resembling elite-school humans more than non-elite ones. He argues this is not explained by provenance-based trust, citing Slocum et al 2025 showing provenance doesn't matter for belief uptake from finetuning. tenobrus jokes this is 'figuring out which ivy league school Claude went to.'
More from Research
- Multi-agent RL post-training fights LLM mode collapse and boosts response diversity — natashajaques · 2026-09-16
- World Models Won't Get Us to AGI — Continual Learning Is the Missing Piece, and It's Hard — Intelligent-Cream-14 · 2026-09-16
- Philosopher adds appendix arguing we can be confident today's LLMs are not conscious — AnnaCiaunica · 2026-09-16
- Princeton builds an erasable, light-programmed ultrathin semiconductor a few molecules thick — MengdiWang10 · 2026-09-16
- Riemannian Metric Matching: amortized diffusion geometry up to 400x faster (ICML Oral) — allen_ai · 2026-09-16
- Scientific Contribution Graph: 6M contributions from 655k papers released — mmbronstein · 2026-09-16