Freeform Preference Learning Defines Robot Rewards in Natural Language

Marcel Torne, Chelsea Finn and colleagues propose freeform preference learning, which defines reward axes via natural language, boosting robotic manipulation performance by 38 percentage points over sparse binary or trajectory preference signals.

2026-09-09 ~ 2026-09-10 · 2 related posts

• Freeform preference learning lets robots take natural-language reward axes over binary feedback — chris_j_paxton · 2026-09-09
• Freeform Preference Learning Boosts Robot Manipulation Over Baselines by 38 Percentage Points — StanfordAILab · 2026-09-10