&Tutor Trains Against Persimmon Human-Learner Model; Swapping It Kills the Gains
gharik · x · 2026-10-31
Humansand argues better models of people lead to better AI for people. Their tutoring system &Tutor learns to teach through interactions with Persimmon, a model of human learners — and the training gains cannot be reproduced when Persimmon is replaced with another learner model. In extensive evals, Persimmon behaves much more like real human learners across conversational moves, affect, and error rates, showing that simulator fidelity can be decisive for agent training.
Related event: Humans& Launches &Tutor: Model the Learner First, Then Train the AI Tutor(13 posts)→
More from AGI Musings
- Reddit stock plunges 42% in 8 trading days as CEO says AI mode no longer drives traffic — SumitGup · 2026-10-31
- HubSpot co-founder: startups die from suicide, not homicide — competition isn't the real killer — nathanbenaich · 2026-10-31
- Grady Booch clashes with AI coding advocates: engineers aren't there to help LLMs — Grady_Booch · 2026-10-31
- Forethought: US has 9x China's AI compute, can afford to pace the frontier — willmacaskill · 2026-10-31
- Asimov predicted machines would take creative work in 1964 — it happened — larabyeol · 2026-10-31
- Forethought: first movers could lock rivals out of space for $49-73B premium — willmacaskill · 2026-10-31