Yoav Goldberg on Whether RL Can Teach Models Factual Knowledge

Yoav Goldberg argues RL can help agents acquire factual knowledge, citing examples like learning to win games via RL. He also pressed Google researchers on whether they, unlike OpenAI, avoid using billions of RL trajectories for knowledge learning.

2026-10-04 ~ 2026-10-04 · 2 related posts

• Yoav Goldberg: RL tasks like winning games or replicating bibliographies teach models facts — yoavgo · 2026-10-04
• Yoav Goldberg asks if Google skips billions of RL trajectories unlike OpenAI — yoavgo · 2026-10-04