BEHAVIOR 1K MP Collected — turning on radio Combined dataset for BEHAVIOR 1K task 0 ( turning on radio ): 1154 success demos + 846 failure demos collected by a hybrid motion planner + X VLA policy pipeline on instances 301–700 (private test set, 400 instances × 5 episodes) 200 success demos from the original BEHAVIOR 1K teleoperated dataset ( behavior 1k/2025 challenge demos ), merged into success/ Total: 1354 success + 846 failure = 2200 episodes (~157 GB). Layout How to use the dataset Download Or from Python: Files are stored individually and match the BEHAVIOR 1K canonical layout — no extraction step required. demo id ranges Source demo id range Count HF original (instances 1–300, ep 0) 10..3000 (step 10) 200 MP collected (instances 301–700, ep 0..4) 3010..7004 1154 success + 846 failure Coverage by artifact parquet, annotations : all 2200 episodes meta + bddl transitions, hdf5, depth videos, rgb.external video : only MP collected (2000 episodes); HF original demos lack these rgb.head / rgb.left wrist / rgb.right wrist videos : all 2200 episodes phase segments : only MP collected (2000 episodes); see below Per episode skill annotation Each annotations/.../episode XXXXXXXX.json…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy