Planner distillation makes agents follow belief-dependent rules 18x more often than RL
StephenLCasper · x · 2026-11-12
philljkc shares new work on getting agents to follow belief-dependent rules under uncertainty: a planner distillation method for POMDPs whose trained policies comply 18x as often as RL policies in a multimodal task with state and goal uncertainty.
More from Research
- Three neurons control Gemma 4's caring vs contempt, exported as <70KB llama.cpp vectors — New_Alps_5655 · 2026-11-12
- MIT's differentiable acoustic simulator reverse-engineers vocal tract from audio — maier_ak · 2026-11-12
- Google Ships Gemini Decision 1: Single-Pass Decision Model at $0.10 per 1M Tokens — Latent Space · 2026-11-12
- MIT Trains Climate Emulators on Synthetic Data, Matching Accuracy with a Fraction of the Data — MIT News AI · 2026-11-12
- CoRL 2026 workshop makes the case for continually self-improving robots — chris_j_paxton · 2026-11-12
- GPT-4 tutoring lifted practice scores 48% but left students 17% worse on exams — tallmetommy · 2026-11-12