BEHAVIOR-1K · Compost Q=0 vs Q=1

Choose a seed to inspect its camera videos and evaluation artifacts.

learned policy q score 0

π0.5 corrective LoRA · cumulative +500 · Q=0

composting_waste · train instance 293 · checkpoint compost_corrective_basegrip8_total500/299

released expert q score 1

Released expert episode 18194 · Q=1

composting_waste · train instance 293 · released demonstration episode 18194

Video

Esc or click outside to close