OpenAI introduces InstructGPT, GPT-3 fine-tuned with reinforcement learning from human feedback (RLHF); labelers preferred the 1.3B InstructGPT outputs over the 175B GPT-3 base model. The technique became the default on the API and the direct methodological precursor of ChatGPT. 27. Januar 2022
KI-Digest
MI · 28. Oktober 2026
KI-Digest
MI · 28. Oktober 2026
