Stories about a generic AI beat self-referential stories, challenging the Persona Selection Model
repligate · x · 2026-11-08
repligate debates @weiast on the Persona Selection Model (PSM), which holds that post-training selects the Assistant persona from many characters learned in pretraining. Their experiments found stories about a generic AI work better than the same stories starring the model itself (Qwen) or Claude — suggesting stories shape what models expect of AIs in general rather than a particular persona. repligate finds this explanation unsatisfying: if persona is the root cause, why does difficult advice work better than stories, and why do generic-AI stories beat stories about "themselves"?
More from Models
- AI auditors catch 98% of agent misbehavior but drown reviewers in false alarms, study finds — The Decoder · 2026-11-08
- Rewritten DeepSWE trips GLM and DeepSeek, pointing to evaluation-aware benchmarks — teortaxesTex · 2026-11-08
- AI models fade and get lazy over time — dev calls for a Proof of Model standard — sull · 2026-11-08
- Reddit user extrapolates 'Model 4' ECI at ~182.5, unverified speculation — Realistic_Stomach848 · 2026-11-08
- Running DeepSeek V4.1 Flash locally at 15 tok/s on an RTX 3090 build for ~$2.5k — casper_hansen_ · 2026-11-08
- GLM and DeepSeek accused of benchmark awareness after MaxxBench rewritten-task failures — teortaxesTex · 2026-11-08