File size: 34,648 Bytes
d9be7c0 948bb27 d9be7c0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 | {
"title": "Ropedia Xperience-10M Unified 20-Task Suite",
"status": "pass",
"generated_at_utc": "2026-06-16T06:26:17+00:00",
"task_count": 20,
"task_count_breakdown": {
"original_public_sample_tasks": 12,
"additional_public_sample_tasks": 8,
"total_unified_tasks": 20
},
"unification_policy": {
"public_framing": "The suite is presented as one 20-task benchmark surface. Tasks 1-12 are the original public-sample tasks; tasks 13-20 are additional sample-supported tasks that use the same window/split/baseline contract.",
"legacy_path_note": "The directory and file name tier2_task_suite are retained only for backward-compatible artifact links; they are not a separate public benchmark tier."
},
"dataset_scope": {
"sample_episode_count": 1,
"annotation": "data/sample/xperience-10m-sample/annotation.hdf5",
"num_frames": 5821,
"num_windows": 1161,
"feature_dim": 8546,
"window_frames": 20,
"stride_frames": 5,
"split_policy": "single_episode_chronological_70_30",
"raw_hdf5_required_for_tasks_13_20_regeneration": true,
"raw_data_redistributed": false
},
"setup_alignment": {
"same_window_unit": "20-frame aligned windows",
"same_stride": "5 frames",
"same_feature_manifest": "results/episode_task_suite/feature_manifest.json",
"same_shared_tensor": "results/episode_task_suite/shared_windows.npz",
"same_split": "chronological 70/30 train/test split within the public sample episode",
"same_baseline_pattern": "minimal interpretable heads plus compact neural MLP heads",
"same_leakage_policy": "Target-side future, contact, object, caption, relation, and interaction signals are excluded from inputs unless language is explicitly the query."
},
"source_files": [
"docs/data/summary_metrics.json",
"docs/data/task_walkthroughs.json",
"docs/data/tier2_task_suite.json",
"results/episode_task_suite/summary_report.json",
"results/episode_task_suite/tier2_task_suite/tier2_task_suite_results.json",
"results/episode_task_suite/windows.csv",
"results/episode_task_suite/feature_manifest.json"
],
"tasks": [
{
"task_id": "timeline_action",
"task_display_name": "Action Recognition",
"research_name": "Egocentric Action Recognition",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "supervised",
"architecture_family": "multiclass classifier",
"primary_direction": "C. Egocentric Vision & Interaction",
"input": "One 20-frame window represented by the current feature vector: video/audio/depth summaries, pose, SLAM/camera pose, motion capture, IMU, calibration, and language-derived context.",
"input_short": "20-frame multimodal window",
"process": "window features -> action label builder -> classifier",
"output": "A single action class for the current window.",
"output_short": "current action class",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.05,
"neural_primary_metric": 0.014814814814814814,
"counts": {
"num_windows": 1144,
"num_eval_windows": 343,
"num_train_windows": 801,
"num_test_windows": 343,
"num_classes": 18
},
"meaning": "Recognize the current manipulation action from synchronized visual, motion, inertial, pose, and annotation context.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/timeline_action.md",
"minimal_metrics": "results/episode_task_suite/timeline_action/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/timeline_action/metrics.json"
},
"task_number": 1,
"suite_label": "Task 01"
},
{
"task_id": "timeline_subtask",
"task_display_name": "Procedure Step Recognition",
"research_name": "Temporal Subtask Recognition",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "supervised",
"architecture_family": "multiclass classifier",
"primary_direction": "C. Egocentric Vision & Interaction",
"input": "The same all-modality window vector used by action recognition.",
"input_short": "20-frame multimodal window",
"process": "window features -> subtask label builder -> classifier",
"output": "A single subtask label for the current window.",
"output_short": "current procedure step",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.05056355513846935,
"neural_primary_metric": 0.02810810810810811,
"counts": {
"num_windows": 1147,
"num_eval_windows": 344,
"num_train_windows": 803,
"num_test_windows": 344,
"num_classes": 14
},
"meaning": "Recognize the broader activity stage so fine actions become a readable procedure timeline.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/timeline_subtask.md",
"minimal_metrics": "results/episode_task_suite/timeline_subtask/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/timeline_subtask/metrics.json"
},
"task_number": 2,
"suite_label": "Task 02"
},
{
"task_id": "transition_detection",
"task_display_name": "Action Boundary Detection",
"research_name": "Temporal Action Segmentation",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "diagnostic",
"architecture_family": "binary classifier",
"primary_direction": "C. Egocentric Vision & Interaction",
"input": "One all-modality window vector plus labels derived from action-change timestamps.",
"input_short": "current window with boundary target",
"process": "action changes -> boundary labels -> binary classifier",
"output": "A binary label: boundary or steady.",
"output_short": "boundary or steady",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.6118237590630229,
"neural_primary_metric": 0.5862068965517241,
"counts": {
"num_windows": 1161,
"num_eval_windows": 348,
"num_train_windows": 813,
"num_test_windows": 348,
"num_classes": 2
},
"meaning": "Detect the local moment where the episode changes from one action segment to the next.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/transition_detection.md",
"minimal_metrics": "results/episode_task_suite/transition_detection/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/transition_detection/metrics.json"
},
"task_number": 3,
"suite_label": "Task 03"
},
{
"task_id": "next_action",
"task_display_name": "Next-Action Prediction",
"research_name": "Short-Horizon Intention Prediction",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "supervised",
"architecture_family": "future-label classifier",
"primary_direction": "C. Egocentric Vision & Interaction",
"input": "The current all-modality window vector at time t.",
"input_short": "current window at time t",
"process": "current features -> future label shift -> classifier",
"output": "A single action class for t+20 frames.",
"output_short": "action at t+20 frames",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.05925925925925927,
"neural_primary_metric": 0.04186046511627907,
"counts": {
"num_windows": 1161,
"num_eval_windows": 348,
"num_train_windows": 813,
"num_test_windows": 348,
"num_classes": 18
},
"meaning": "Forecast the near-future action from the current observations only.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/next_action.md",
"minimal_metrics": "results/episode_task_suite/next_action/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/next_action/metrics.json"
},
"task_number": 4,
"suite_label": "Task 04"
},
{
"task_id": "hand_trajectory_forecast",
"task_display_name": "Hand Trajectory Forecasting",
"research_name": "3D Hand Motion Forecasting",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "forecast",
"architecture_family": "continuous regressor",
"primary_direction": "A. Human Modeling & Motion Understanding",
"input": "The current all-modality window vector at time t.",
"input_short": "current multimodal window",
"process": "current features -> future mocap target -> regression head",
"output": "A future trajectory vector for left and right hand joints.",
"output_short": "future hand-joint trajectory",
"metric_key": "mpjpe",
"metric_name": "MPJPE",
"metric_direction": "lower",
"minimal_primary_metric": 0.8646570444107056,
"neural_primary_metric": 0.10785018652677536,
"counts": {
"num_windows": 1159,
"num_train_windows": 811,
"num_test_windows": 348
},
"meaning": "Predict the future 3D left/right hand path from the current multimodal state.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/hand_trajectory_forecast.md",
"minimal_metrics": "results/episode_task_suite/hand_trajectory_forecast/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/hand_trajectory_forecast/metrics.json"
},
"task_number": 5,
"suite_label": "Task 05"
},
{
"task_id": "contact_prediction",
"task_display_name": "Contact State Prediction",
"research_name": "Human-Object Contact Prediction",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "supervised",
"architecture_family": "binary classifier",
"primary_direction": "A. Human Modeling & Motion Understanding",
"input": "Non-contact and non-caption feature blocks, so the answer is not directly leaked from the target labels.",
"input_short": "non-contact, non-caption features",
"process": "feature filter -> contact target -> binary classifier",
"output": "A binary contact label.",
"output_short": "contact or no contact",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 1.0,
"neural_primary_metric": 1.0,
"counts": {
"num_windows": 1161,
"num_eval_windows": 348,
"num_train_windows": 813,
"num_test_windows": 348,
"num_classes": 1
},
"meaning": "Predict whether body or hand contact with the scene is occurring without leaking contact labels.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/contact_prediction.md",
"minimal_metrics": "results/episode_task_suite/contact_prediction/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/contact_prediction/metrics.json"
},
"task_number": 6,
"suite_label": "Task 06"
},
{
"task_id": "object_relevance",
"task_display_name": "Object Relevance Prediction",
"research_name": "Object-Centric Interaction Recognition",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "supervised",
"architecture_family": "multi-label classifier",
"primary_direction": "C. Egocentric Vision & Interaction",
"input": "Non-caption feature blocks, so the model must infer objects from sensors rather than copying the caption words.",
"input_short": "non-caption multimodal features",
"process": "object vocabulary -> multi-hot labels -> sigmoid heads",
"output": "A multi-label object set for the current window.",
"output_short": "relevant object set",
"metric_key": "micro_f1",
"metric_name": "micro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.18034382095361662,
"neural_primary_metric": 0.1679279279279279,
"counts": {
"num_windows": 1161,
"num_train_windows": 813,
"num_test_windows": 348
},
"meaning": "Infer which objects are relevant to the current manipulation window from non-caption features.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/object_relevance.md",
"minimal_metrics": "results/episode_task_suite/object_relevance/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/object_relevance/metrics.json"
},
"task_number": 7,
"suite_label": "Task 07"
},
{
"task_id": "caption_grounding",
"task_display_name": "Language Grounding",
"research_name": "Language-to-Moment Grounding",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "retrieval",
"architecture_family": "retrieval ranker",
"primary_direction": "C. Egocentric Vision & Interaction",
"input": "Caption/object/interaction query features and a set of candidate sensor-window features.",
"input_short": "text-like query and candidate windows",
"process": "query features -> candidate index -> cosine ranker",
"output": "A ranked list of windows, with the correct matching window ideally near rank 1.",
"output_short": "ranked matching moments",
"metric_key": "mrr",
"metric_name": "MRR",
"metric_direction": "higher",
"minimal_primary_metric": 0.016023479050338015,
"neural_primary_metric": 0.01684125567132316,
"counts": {
"num_queries": 348,
"num_train_windows": 813,
"num_test_windows": 348
},
"meaning": "Retrieve the matching time window for an annotation-derived text query.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/caption_grounding.md",
"minimal_metrics": "results/episode_task_suite/caption_grounding/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/caption_grounding/metrics.json"
},
"task_number": 8,
"suite_label": "Task 08"
},
{
"task_id": "cross_modal_retrieval",
"task_display_name": "Cross-Modal Retrieval",
"research_name": "Multimodal Representation Retrieval",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "retrieval",
"architecture_family": "two-tower retrieval head",
"primary_direction": "D. Scene Reconstruction & World Modeling",
"input": "Query side: motion, IMU, and camera/pose features. Candidate side: depth and video features.",
"input_short": "motion/IMU/pose query; depth/video candidates",
"process": "modality split -> projection -> nearest-neighbor ranker",
"output": "A ranked list of candidate depth/video windows.",
"output_short": "ranked visual windows",
"metric_key": "mrr",
"metric_name": "MRR",
"metric_direction": "higher",
"minimal_primary_metric": 0.26925966892956127,
"neural_primary_metric": 0.1299971898648288,
"counts": {
"num_queries": 348,
"num_train_windows": 813,
"num_test_windows": 348
},
"meaning": "Use motion, IMU, and camera-pose signals to retrieve the matching depth/video window.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/cross_modal_retrieval.md",
"minimal_metrics": "results/episode_task_suite/cross_modal_retrieval/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/cross_modal_retrieval/metrics.json"
},
"task_number": 9,
"suite_label": "Task 09"
},
{
"task_id": "modality_reconstruction",
"task_display_name": "Cross-Modal Reconstruction",
"research_name": "Modality Feature Reconstruction",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "forecast",
"architecture_family": "feature regressor",
"primary_direction": "B. 3D/4D Reconstruction & Neural Rendering",
"input": "Motion, IMU, and camera/pose features as input; depth/video features as the regression target.",
"input_short": "motion, IMU, and camera/pose features",
"process": "source-target split -> scaler -> regression head",
"output": "A reconstructed depth/video feature vector.",
"output_short": "reconstructed depth/video vector",
"metric_key": "r2",
"metric_name": "R2",
"metric_direction": "higher",
"minimal_primary_metric": -0.015271898913936655,
"neural_primary_metric": -0.010171410134180991,
"counts": {
"num_train_windows": 813,
"num_test_windows": 348
},
"meaning": "Predict compressed depth/video feature vectors from motion, IMU, and camera-pose features.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/modality_reconstruction.md",
"minimal_metrics": "results/episode_task_suite/modality_reconstruction/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/modality_reconstruction/metrics.json"
},
"task_number": 10,
"suite_label": "Task 10"
},
{
"task_id": "temporal_order",
"task_display_name": "Temporal Order Verification",
"research_name": "Temporal Order Verification",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "diagnostic",
"architecture_family": "pairwise classifier",
"primary_direction": "D. Scene Reconstruction & World Modeling",
"input": "A pair of adjacent window vectors, plus their difference vector.",
"input_short": "two adjacent windows plus difference vector",
"process": "pair builder -> feature combiner -> binary classifier",
"output": "A binary label: correct order or reversed order.",
"output_short": "correct or reversed",
"metric_key": "f1",
"metric_name": "F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.5399515738498789,
"neural_primary_metric": 0.8520179372197308,
"counts": {
"num_samples": 2320,
"num_train_samples": 1624,
"num_test_samples": 696
},
"meaning": "Tell whether two neighboring windows are in chronological order or reversed.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/temporal_order.md",
"minimal_metrics": "results/episode_task_suite/temporal_order/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/temporal_order/metrics.json"
},
"task_number": 11,
"suite_label": "Task 11"
},
{
"task_id": "misalignment_detection",
"task_display_name": "Multimodal Synchronization Detection",
"research_name": "Cross-Modal Misalignment Detection",
"origin": "original_public_sample_tasks",
"origin_count_label": "original task",
"family": "diagnostic",
"architecture_family": "pairwise classifier",
"primary_direction": "B. 3D/4D Reconstruction & Neural Rendering",
"input": "A motion-side feature group and a visual/depth-side feature group, either aligned or artificially shifted.",
"input_short": "motion-side and visual/depth-side feature groups",
"process": "aligned/shifted pairs -> feature combiner -> binary classifier",
"output": "A binary label: aligned or shifted.",
"output_short": "aligned or shifted",
"metric_key": "f1",
"metric_name": "F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.5051698670605613,
"neural_primary_metric": 0.7152682255845944,
"counts": {
"num_samples": 2306,
"num_train_samples": 1614,
"num_test_samples": 692
},
"meaning": "Detect whether motion and visual/depth streams have been artificially shifted out of sync.",
"artifact_sources": {
"walkthrough": "results/episode_task_suite/task_walkthroughs/misalignment_detection.md",
"minimal_metrics": "results/episode_task_suite/misalignment_detection/metrics.json",
"neural_metrics": "results/episode_task_suite/neural_mlp/misalignment_detection/metrics.json"
},
"task_number": 12,
"suite_label": "Task 12"
},
{
"task_id": "long_horizon_next_action",
"task_display_name": "Long-Horizon Next-Action Forecasting",
"research_name": "Long-Horizon Next-Action Forecasting",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "classification",
"architecture_family": "minimal_softmax",
"primary_direction": "sample-supported extension",
"input": "Current 20-frame non-caption multimodal window.",
"input_short": "Current 20-frame non-caption multimodal window.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "Action label five seconds later.",
"output_short": "Action label five seconds later.",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.07499999999999998,
"neural_primary_metric": 0.06545454545454546,
"counts": {
"num_windows": 1073,
"num_eval_windows": 322,
"num_train_windows": 751,
"num_test_windows": 322,
"num_classes": 18
},
"meaning": "Tests whether the current state carries enough procedure context to forecast beyond the one-second core next-action task.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/long_horizon_next_action/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/long_horizon_next_action/metrics.json"
},
"task_number": 13,
"suite_label": "Task 13"
},
{
"task_id": "next_subtask_forecast",
"task_display_name": "Long-Horizon Next-Subtask Forecasting",
"research_name": "Long-Horizon Next-Subtask Forecasting",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "classification",
"architecture_family": "minimal_softmax",
"primary_direction": "sample-supported extension",
"input": "Current 20-frame non-caption multimodal window.",
"input_short": "Current 20-frame non-caption multimodal window.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "Procedure subtask label five seconds later.",
"output_short": "Procedure subtask label five seconds later.",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.04545454545454545,
"neural_primary_metric": 0.050724637681159424,
"counts": {
"num_windows": 1141,
"num_eval_windows": 342,
"num_train_windows": 799,
"num_test_windows": 342,
"num_classes": 14
},
"meaning": "Moves from immediate action anticipation to higher-level procedure-state prediction.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/next_subtask_forecast/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/next_subtask_forecast/metrics.json"
},
"task_number": 14,
"suite_label": "Task 14"
},
{
"task_id": "interaction_text_prediction",
"task_display_name": "Interaction Text Prediction",
"research_name": "Interaction Text Prediction",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "classification",
"architecture_family": "minimal_softmax",
"primary_direction": "sample-supported extension",
"input": "Current 20-frame sensor window with caption-text features removed.",
"input_short": "Current 20-frame sensor window with caption-text features removed.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "Raw annotation interaction phrase for the same window.",
"output_short": "Raw annotation interaction phrase for the same window.",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.04444444444444444,
"neural_primary_metric": 0.0380952380952381,
"counts": {
"num_windows": 192,
"num_eval_windows": 58,
"num_train_windows": 134,
"num_test_windows": 58,
"num_classes": 46
},
"meaning": "Uses the raw caption JSON interaction field as a language target instead of only the hashed text feature.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/interaction_text_prediction/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/interaction_text_prediction/metrics.json"
},
"task_number": 15,
"suite_label": "Task 15"
},
{
"task_id": "action_object_relation",
"task_display_name": "Action-Object Relation Prediction",
"research_name": "Action-Object Relation Prediction",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "classification",
"architecture_family": "minimal_softmax",
"primary_direction": "sample-supported extension",
"input": "Current 20-frame sensor window with caption-text features removed.",
"input_short": "Current 20-frame sensor window with caption-text features removed.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "Joint action plus active object-set relation.",
"output_short": "Joint action plus active object-set relation.",
"metric_key": "macro_f1",
"metric_name": "macro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.0,
"neural_primary_metric": 0.0,
"counts": {
"num_windows": 178,
"num_eval_windows": 53,
"num_train_windows": 125,
"num_test_windows": 53,
"num_classes": 42
},
"meaning": "Evaluates whether a model can bind what action is happening to which objects are involved.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/action_object_relation/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/action_object_relation/metrics.json"
},
"task_number": 16,
"suite_label": "Task 16"
},
{
"task_id": "object_set_forecast",
"task_display_name": "Future Object-Set Forecasting",
"research_name": "Future Object-Set Forecasting",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "multi_label",
"architecture_family": "minimal_ridge_multilabel",
"primary_direction": "sample-supported extension",
"input": "Current 20-frame sensor window with caption-text features removed.",
"input_short": "Current 20-frame sensor window with caption-text features removed.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "Object set active five seconds later.",
"output_short": "Object set active five seconds later.",
"metric_key": "micro_f1",
"metric_name": "micro-F1",
"metric_direction": "higher",
"minimal_primary_metric": 0.16939890710382516,
"neural_primary_metric": 0.19718309859154928,
"counts": {
"num_windows": 188,
"num_train_windows": 132,
"num_test_windows": 56
},
"meaning": "Predicts which objects will become relevant soon, not only which objects are relevant now.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/object_set_forecast/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/object_set_forecast/metrics.json"
},
"task_number": 17,
"suite_label": "Task 17"
},
{
"task_id": "imu_to_hand_pose",
"task_display_name": "IMU-to-Hand Pose Reconstruction",
"research_name": "IMU-to-Hand Pose Reconstruction",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "regression",
"architecture_family": "minimal_ridge_regression",
"primary_direction": "sample-supported extension",
"input": "Current IMU acceleration/gyroscope feature block only.",
"input_short": "Current IMU acceleration/gyroscope feature block only.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "Current left/right hand joint feature blocks.",
"output_short": "Current left/right hand joint feature blocks.",
"metric_key": "mae",
"metric_name": "MAE",
"metric_direction": "lower",
"minimal_primary_metric": 0.042049407958984375,
"neural_primary_metric": 0.042562149465084076,
"counts": {
"num_windows": 1161,
"num_train_windows": 813,
"num_test_windows": 348
},
"meaning": "A sensor-bridge probe for how much hand configuration can be recovered from inertial motion alone.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/imu_to_hand_pose/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/imu_to_hand_pose/metrics.json"
},
"task_number": 18,
"suite_label": "Task 18"
},
{
"task_id": "camera_view_sync_retrieval",
"task_display_name": "Camera-View Synchronization Retrieval",
"research_name": "Camera-View Synchronization Retrieval",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "retrieval",
"architecture_family": "minimal_ridge_projection_cosine_retrieval",
"primary_direction": "sample-supported extension",
"input": "Fisheye camera-1 feature query projected into fisheye camera-3 feature space.",
"input_short": "Fisheye camera-1 feature query projected into fisheye camera-3 feature space.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "The synchronized held-out camera-3 window.",
"output_short": "The synchronized held-out camera-3 window.",
"metric_key": "mrr",
"metric_name": "MRR",
"metric_direction": "higher",
"minimal_primary_metric": 0.4943004846572876,
"neural_primary_metric": 0.24086658656597137,
"counts": {
"num_train_windows": 813,
"num_test_windows": 348
},
"meaning": "Stress-tests multi-camera time alignment beyond the core cross-modal retrieval task.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/camera_view_sync_retrieval/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/camera_view_sync_retrieval/metrics.json"
},
"task_number": 19,
"suite_label": "Task 19"
},
{
"task_id": "time_to_transition",
"task_display_name": "Time-to-Next-Transition Regression",
"research_name": "Time-to-Next-Transition Regression",
"origin": "additional_public_sample_tasks",
"origin_count_label": "additional task",
"family": "regression",
"architecture_family": "minimal_ridge_regression",
"primary_direction": "sample-supported extension",
"input": "Current 20-frame non-caption multimodal window.",
"input_short": "Current 20-frame non-caption multimodal window.",
"process": "shared window features -> task-specific target builder -> minimal/neural head",
"output": "Frames until the next action-label boundary, capped at 200 frames.",
"output_short": "Frames until the next action-label boundary, capped at 200 frames.",
"metric_key": "mae",
"metric_name": "MAE frames",
"metric_direction": "lower",
"minimal_primary_metric": 10.53735637664795,
"neural_primary_metric": 10.55449390411377,
"counts": {
"num_windows": 1161,
"num_train_windows": 813,
"num_test_windows": 348
},
"meaning": "Turns boundary detection into a continuous timing estimate for procedural control.",
"artifact_sources": {
"legacy_result_directory": "results/episode_task_suite/tier2_task_suite/",
"minimal_metrics": "results/episode_task_suite/tier2_task_suite/time_to_transition/metrics.json",
"neural_metrics": "results/episode_task_suite/tier2_task_suite/neural_mlp/time_to_transition/metrics.json"
},
"task_number": 20,
"suite_label": "Task 20"
}
]
}
|