{
 "stem": "eval__fontaine_molmo2_er_60k_ddp4__step_060000__panel_curated_v0_k4l2_events",
 "frames_scored": 8987,
 "unparsed_generations": 0,
 "presence_accuracy": 0.8567931456548348,
 "banked_event_acc": 0.8582396795371091,
 "oracle_tolerance": 0.003,
 "exact_string_match_on_gt_event_frames": 0.03580786026200873,
 "buckets": {
  "both_none": 7238,
  "miss": 683,
  "false_alarm": 604,
  "hit": 333,
  "class_swap": 129
 },
 "classes": [
  "none",
  "episode_marker",
  "human",
  "idle",
  "blur",
  "occlusion",
  "miss_fail",
  "release_place",
  "drop_slip",
  "grasp_pickup",
  "contact_collision",
  "camera_view",
  "other"
 ],
 "matrix": {
  "none": {
   "none": 7238,
   "episode_marker": 2,
   "human": 9,
   "idle": 255,
   "blur": 49,
   "occlusion": 54,
   "miss_fail": 1,
   "release_place": 50,
   "drop_slip": 0,
   "grasp_pickup": 11,
   "contact_collision": 4,
   "camera_view": 3,
   "other": 166
  },
  "episode_marker": {
   "none": 19,
   "episode_marker": 3,
   "human": 1,
   "idle": 4,
   "blur": 1,
   "occlusion": 0,
   "miss_fail": 0,
   "release_place": 0,
   "drop_slip": 0,
   "grasp_pickup": 0,
   "contact_collision": 0,
   "camera_view": 1,
   "other": 4
  },
  "human": {
   "none": 18,
   "episode_marker": 1,
   "human": 31,
   "idle": 1,
   "blur": 0,
   "occlusion": 0,
   "miss_fail": 0,
   "release_place": 0,
   "drop_slip": 0,
   "grasp_pickup": 0,
   "contact_collision": 0,
   "camera_view": 0,
   "other": 2
  },
  "idle": {
   "none": 219,
   "episode_marker": 1,
   "human": 0,
   "idle": 157,
   "blur": 1,
   "occlusion": 0,
   "miss_fail": 0,
   "release_place": 1,
   "drop_slip": 0,
   "grasp_pickup": 0,
   "contact_collision": 0,
   "camera_view": 0,
   "other": 15
  },
  "blur": {
   "none": 90,
   "episode_marker": 0,
   "human": 0,
   "idle": 0,
   "blur": 57,
   "occlusion": 7,
   "miss_fail": 0,
   "release_place": 2,
   "drop_slip": 0,
   "grasp_pickup": 0,
   "contact_collision": 2,
   "camera_view": 2,
   "other": 3
  },
  "occlusion": {
   "none": 116,
   "episode_marker": 0,
   "human": 0,
   "idle": 0,
   "blur": 4,
   "occlusion": 37,
   "miss_fail": 1,
   "release_place": 2,
   "drop_slip": 0,
   "grasp_pickup": 0,
   "contact_collision": 2,
   "camera_view": 0,
   "other": 7
  },
  "miss_fail": {
   "none": 21,
   "episode_marker": 0,
   "human": 0,
   "idle": 0,
   "blur": 0,
   "occlusion": 0,
   "miss_fail": 0,
   "release_place": 0,
   "drop_slip": 0,
   "grasp_pickup": 2,
   "contact_collision": 1,
   "camera_view": 0,
   "other": 3
  },
  "release_place": {
   "none": 71,
   "episode_marker": 0,
   "human": 0,
   "idle": 0,
   "blur": 2,
   "occlusion": 2,
   "miss_fail": 1,
   "release_place": 24,
   "drop_slip": 0,
   "grasp_pickup": 0,
   "contact_collision": 0,
   "camera_view": 0,
   "other": 3
  },
  "drop_slip": {
   "none": 2,
   "episode_marker": 0,
   "human": 0,
   "idle": 2,
   "blur": 1,
   "occlusion": 0,
   "miss_fail": 0,
   "release_place": 0,
   "drop_slip": 0,
   "grasp_pickup": 0,
   "contact_collision": 0,
   "camera_view": 0,
   "other": 0
  },
  "grasp_pickup": {
   "none": 12,
   "episode_marker": 0,
   "human": 0,
   "idle": 0,
   "blur": 0,
   "occlusion": 1,
   "miss_fail": 1,
   "release_place": 0,
   "drop_slip": 0,
   "grasp_pickup": 6,
   "contact_collision": 0,
   "camera_view": 0,
   "other": 5
  },
  "contact_collision": {
   "none": 12,
   "episode_marker": 0,
   "human": 0,
   "idle": 0,
   "blur": 2,
   "occlusion": 1,
   "miss_fail": 4,
   "release_place": 1,
   "drop_slip": 0,
   "grasp_pickup": 1,
   "contact_collision": 2,
   "camera_view": 0,
   "other": 2
  },
  "camera_view": {
   "none": 30,
   "episode_marker": 0,
   "human": 0,
   "idle": 2,
   "blur": 1,
   "occlusion": 2,
   "miss_fail": 0,
   "release_place": 1,
   "drop_slip": 0,
   "grasp_pickup": 1,
   "contact_collision": 0,
   "camera_view": 3,
   "other": 7
  },
  "other": {
   "none": 73,
   "episode_marker": 1,
   "human": 0,
   "idle": 3,
   "blur": 0,
   "occlusion": 1,
   "miss_fail": 0,
   "release_place": 5,
   "drop_slip": 0,
   "grasp_pickup": 4,
   "contact_collision": 1,
   "camera_view": 0,
   "other": 13
  }
 },
 "per_class": {
  "none": {
   "gt_frames": 7842,
   "model_frames": 7921,
   "precision": 0.913773513445272,
   "recall": 0.9229788319306299
  },
  "episode_marker": {
   "gt_frames": 33,
   "model_frames": 8,
   "precision": 0.375,
   "recall": 0.09090909090909091
  },
  "human": {
   "gt_frames": 53,
   "model_frames": 41,
   "precision": 0.7560975609756098,
   "recall": 0.5849056603773585
  },
  "idle": {
   "gt_frames": 394,
   "model_frames": 424,
   "precision": 0.37028301886792453,
   "recall": 0.39847715736040606
  },
  "blur": {
   "gt_frames": 163,
   "model_frames": 118,
   "precision": 0.4830508474576271,
   "recall": 0.3496932515337423
  },
  "occlusion": {
   "gt_frames": 169,
   "model_frames": 105,
   "precision": 0.3523809523809524,
   "recall": 0.21893491124260356
  },
  "miss_fail": {
   "gt_frames": 27,
   "model_frames": 8,
   "precision": 0.0,
   "recall": 0.0
  },
  "release_place": {
   "gt_frames": 103,
   "model_frames": 86,
   "precision": 0.27906976744186046,
   "recall": 0.23300970873786409
  },
  "drop_slip": {
   "gt_frames": 5,
   "model_frames": 0,
   "precision": null,
   "recall": 0.0
  },
  "grasp_pickup": {
   "gt_frames": 25,
   "model_frames": 25,
   "precision": 0.24,
   "recall": 0.24
  },
  "contact_collision": {
   "gt_frames": 25,
   "model_frames": 12,
   "precision": 0.16666666666666666,
   "recall": 0.08
  },
  "camera_view": {
   "gt_frames": 47,
   "model_frames": 9,
   "precision": 0.3333333333333333,
   "recall": 0.06382978723404255
  },
  "other": {
   "gt_frames": 101,
   "model_frames": 230,
   "precision": 0.05652173913043478,
   "recall": 0.12871287128712872
  }
 },
 "other_residue": {
  "gt_top": [
   [
    "board state appears unchanged from the first frame",
    2
   ],
   [
    "arm raised high above the blocks, no approach to any block",
    1
   ],
   [
    "arm retreats to an upright pose with no block held",
    1
   ],
   [
    "cookie appears on the paper (scene set up mid-episode)",
    1
   ],
   [
    "board layout appears unchanged versus frame 1",
    1
   ],
   [
    "final board state looks nearly identical to the start; the move is not clearly verifiable",
    1
   ],
   [
    "arm retracts without any piece having been moved",
    1
   ],
   [
    "no observable change in scene",
    1
   ],
   [
    "arm withdraws without moving any piece",
    1
   ],
   [
    "arm retracts away from the board without moving a piece",
    1
   ],
   [
    "robot parked away from the workspace, board untouched",
    1
   ],
   [
    "arm retracts but board shows no clear piece displacement",
    1
   ],
   [
    "block reaches near corner of baseplate; release not clearly visible",
    1
   ],
   [
    "arm holds still, no further motion",
    1
   ],
   [
    "stick initially sits inside the bowl (pre-reset state)",
    1
   ]
  ],
  "model_top": [
   [
    "object 0,1; gripper 0,1",
    77
   ],
   [
    "object 0; gripper 0,1",
    33
   ],
   [
    "object 0; gripper 0",
    23
   ],
   [
    "object 0,1; gripper 0",
    9
   ],
   [
    "arm returns to home pose with cubes untouched",
    6
   ],
   [
    "bolt now lies on the mat instead of in the dish (scene reset)",
    6
   ],
   [
    "object 0; gripper none",
    5
   ],
   [
    "object 0; gripper 1",
    5
   ],
   [
    "arm returns to home pose with both cubes untouched",
    4
   ],
   [
    "arm returns home with no object moved",
    3
   ],
   [
    "robot arm not yet in frame",
    3
   ],
   [
    "arm returns to home with cubes untouched",
    2
   ],
   [
    "object none; gripper 0",
    2
   ],
   [
    "bolt now lies on the mat instead of on the dish (scene reset)",
    2
   ],
   [
    "bolt now lies on the mat instead of on the disc (scene state changed)",
    2
   ]
  ]
 }
}