RLHF

Concetto Addestramento e inferenza

Reinforcement learning from human feedback — the technique behind ChatGPT's manners.

Stai visualizzando una scheda concettuale: cosa significa questo termine in linguaggio chiaro, e i record che mette in relazione.

RLHF trains a reward model from human preference comparisons, then optimizes the language model against it. Introduced at scale by InstructGPT (2022), it is why assistants follow instructions and refuse harm — and its quirks (sycophancy, reward hacking) drive research into alternatives like constitutional AI and direct preference optimization.

Come si collega al resto dello stack

I collegamenti di questa scheda a colpo d'occhio — clicca un gruppo per raggiungerlo.

Introdotto

InstructGPT — RLHF at Scale (2022)

Basato su

InstructGPT — RLHF at Scale (2022)

Correlati

Constitutional AI (2022) Alignment Post-training (RLHF, RLVR and after)

Fonti

InstructGPT paper ↗

Record revisionato al 2026-01-15 — vedi Metodologia.