{ "run_id": "pact_eval100_pm_easycase_probe_20260521", "title": "PAct Eval100 PartNetMobility easy-case probe", "portal_summary": "Manual probe over low-part-count PartNetMobility samples with VLM QA and interactive 3D previews.", "portal_tags": [ "PAct", "Eval100", "easy-cases", "metrics" ], "timestamp": "2026-05-21T17:30:18+00:00", "selection_policy": "manual benchmark_index selection: 50, 58, 62, 66, 67, 77, 82, 89", "seed": 20260521, "pact_steps": "PAct defaults (ss_steps=25, slat_steps=25)", "pact_output_overrides": { "export_textured_glb": false, "fast_output": false }, "dataset_validation": { "dataset_root": "/data/250010098/PAct-Transporter/datasets/pact_eval100_2_1_coverage_20260521", "num_records": 100, "unique_sample_ids": 100, "by_source": { "ArtVIP": 25, "GAPartNet": 25, "GRScenes": 25, "PartNetMobility": 25 }, "by_category": { "Clock": 4, "Dispenser": 4, "architectural_fixtures": 11, "electronics": 5, "household_fixtures": 13, "household_items": 6, "large_furniture": 4, "major_appliances": 22, "small_appliances": 16, "small_furniture": 9, "storage": 6 }, "by_problem_tag": { "cross_source_domain_gap": 75, "hard_motion": 74, "internal_or_occluded_part_risk": 48, "known_quality_or_conversion_issue": 40, "many_parts": 56, "mixed_joint_types": 37, "non_pm_generalization": 75, "part_decomposition": 100, "prismatic_motion": 46, "revolute_motion": 91, "single_view_conditioning": 100, "thin_structure": 9, "unseen_category": 15 }, "parsed_sdf_count": 100, "preview_count": 100, "issues": [], "ok": true }, "sample_records": [ { "sample_name": "050_PartNetMobility_Clock_Clock_6641", "benchmark_index": 50, "sample_id": "PartNetMobility/Clock/6641", "source_dataset": "PartNetMobility", "category": "Clock", "object_id": "Clock/6641", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 6, "texture_like_meshes": 4, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/050_PartNetMobility_Clock_Clock_6641/050_PartNetMobility_Clock_Clock_6641_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/050_PartNetMobility_Clock_Clock_6641@050_PartNetMobility_Clock_Clock_6641_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.5468537379667018, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 1, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 0.5, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 1, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 1, "status": "measured", "note": "" }, "joint_precision": { "value": 0.0, "status": "measured", "note": "" }, "joint_recall": { "value": 0.0, "status": "measured", "note": "" }, "joint_f1": { "value": 0.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 0.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 90.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.0, "status": "measured", "note": "" }, "joint_limit_error": { "value": 0.0, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 1, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.4531462620332982, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 0.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 1, "gt_part_count": 2, "pred_joint_count": 0, "gt_joint_count": 1, "matches": [], "pred_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 54.68537379667018, "part_decomposition": 50.0, "kinematics": 0.0, "motion_consistency": 0.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 29.687164283684762, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/050_PartNetMobility_Clock_Clock_6641.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "以下是对PAct在给定样本上表现的简短测评和问题发现：\n\n1. **几何完整性差，Part分解不准确：** 生成的三维模型几何结构不完整，存在明显的缺失部分（例如，时钟的背面和侧面缺失），且与原始输入形状差异较大。Part分解也存在问题，GT有2个part，但PAct的part_count_mae=1，说明它只识别出了1个part，未能正确分解。\n2. **关节识别完全失败，运动不可行：** joint_f1=0.000表明PAct未能识别出任何关节，导致无法推断出任何可行的运动。这与GT movable=1相悖，说明模型完全无法理解物体的铰链特性。\n3. **对2D分割的依赖性暴露：** 2D part mask显示只有一个大的蓝色区域，这可能直接导致PAct将整个物体视为一个单一的part，从而无法进行进一步的part分解和关节识别。这暴露了模型对2D分割输入的强依赖性，如果2D分割不准确或过于简化，3D生成结果也会受到严重影响。\n4. **PAct score较低，反映整体性能不佳：** PAct weighted_score=29.69，相对较低，进一步印证了模型在该样本上在几何、part分解和关节识别方面的综合性能不佳。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "以下是对PAct在给定样本上表现的简短测评和问题发现：\n\n1. **几何完整性差，Part分解不准确：** 生成的三维模型几何结构不完整，存在明显的缺失部分（例如，时钟的背面和侧面缺失），且与原始输入形状差异较大。Part分解也存在问题，GT有2个part，但PAct的part_count_mae=1，说明它只识别出了1个part，未能正确分解。\n2. **关节识别完全失败，运动不可行：** joint_f1=0.000表明PAct未能识别出任何关节，导致无法推断出任何可行的运动。这与GT movable=1相悖，说明模型完全无法理解物体的铰链特性。\n3. **对2D分割的依赖性暴露：** 2D part mask显示只有一个大的蓝色区域，这可能直接导致PAct将整个物体视为一个单一的part，从而无法进行进一步的part分解和关节识别。这暴露了模型对2D分割输入的强依赖性，如果2D分割不准确或过于简化，3D生成结果也会受到严重影响。\n4. **PAct score较低，反映整体性能不佳：** PAct weighted_score=29.69，相对较低，进一步印证了模型在该样本上在几何、part分解和关节识别方面的综合性能不佳。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 413, "candidatesTokenCount": 305, "totalTokenCount": 718, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 155 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "bkEPaonOBKmfz7IPjdyl6Q0" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/050_PartNetMobility_Clock_Clock_6641/source_textured.glb", "gt_parts_glb": "viewer_assets/050_PartNetMobility_Clock_Clock_6641/gt_parts.glb", "pact_glb": "viewer_assets/050_PartNetMobility_Clock_Clock_6641/pact_parts.glb" } }, { "sample_name": "058_PartNetMobility_small_appliances_small_appliances_3483", "benchmark_index": 58, "sample_id": "PartNetMobility/small_appliances/3483", "source_dataset": "PartNetMobility", "category": "small_appliances", "object_id": "small_appliances/3483", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 46, "texture_like_meshes": 0, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/058_PartNetMobility_small_appliances_small_appliances_3483/058_PartNetMobility_small_appliances_small_appliances_3483_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/058_PartNetMobility_small_appliances_small_appliances_3483@058_PartNetMobility_small_appliances_small_appliances_3483_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.6132554244474869, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 0, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 0, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 0, "status": "measured", "note": "" }, "joint_precision": { "value": 0.0, "status": "measured", "note": "" }, "joint_recall": { "value": 0.0, "status": "measured", "note": "" }, "joint_f1": { "value": 0.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 0.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 90.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.0, "status": "measured", "note": "" }, "joint_limit_error": { "value": 0.0, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 1, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.38674457555251307, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 0.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 2, "gt_part_count": 2, "pred_joint_count": 1, "gt_joint_count": 1, "matches": [ { "type_mismatch": 1.0, "axis_error_deg": 90.0, "origin_error": 0.6068174651920257, "limit_error": 71.8715087890625, "accepted": false } ], "pred_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 61.32554244474869, "part_decomposition": 100.0, "kinematics": 0.0, "motion_consistency": 0.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 41.261927754434005, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/058_PartNetMobility_small_appliances_small_appliances_3483.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "以下是对PAct在样本3483上的简短测评和问题发现：\n\n1. **几何完整性良好，但Part分解不准确：** PAct生成的几何形状与输入模型高度一致，但未能正确识别出GT中可移动的部件（GT movable=1），而是将整个物体视为一个整体（2D mask显示只有一个蓝色大块和一个小红色块）。\n2. **关节数量/类型识别失败：** PAct的joint_f1为0.000，表明它完全未能识别出任何关节，也因此无法推断出关节类型。这直接导致了运动可行性评估的失败。\n3. **对2D分割的依赖性暴露：** 2D part mask仅将顶部的小把手（红色）与主体（蓝色）分开，而GT parts=2且GT movable=1，暗示主体内部应有可移动部件（例如烤面包机的槽）。PAct的输出似乎过于依赖这个粗粒度的2D分割，未能进行更深层次的语义理解和部件分解。\n4. **运动可行性无法评估：** 由于未能识别出可移动部件和关节，PAct无法为该物体生成任何有意义的运动。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "以下是对PAct在样本3483上的简短测评和问题发现：\n\n1. **几何完整性良好，但Part分解不准确：** PAct生成的几何形状与输入模型高度一致，但未能正确识别出GT中可移动的部件（GT movable=1），而是将整个物体视为一个整体（2D mask显示只有一个蓝色大块和一个小红色块）。\n2. **关节数量/类型识别失败：** PAct的joint_f1为0.000，表明它完全未能识别出任何关节，也因此无法推断出关节类型。这直接导致了运动可行性评估的失败。\n3. **对2D分割的依赖性暴露：** 2D part mask仅将顶部的小把手（红色）与主体（蓝色）分开，而GT parts=2且GT movable=1，暗示主体内部应有可移动部件（例如烤面包机的槽）。PAct的输出似乎过于依赖这个粗粒度的2D分割，未能进行更深层次的语义理解和部件分解。\n4. **运动可行性无法评估：** 由于未能识别出可移动部件和关节，PAct无法为该物体生成任何有意义的运动。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 419, "candidatesTokenCount": 264, "totalTokenCount": 683, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 161 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "c0EPasH7AcGez7IP0bqnqA8" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/058_PartNetMobility_small_appliances_small_appliances_3483/source_textured.glb", "gt_parts_glb": "viewer_assets/058_PartNetMobility_small_appliances_small_appliances_3483/gt_parts.glb", "pact_glb": "viewer_assets/058_PartNetMobility_small_appliances_small_appliances_3483/pact_parts.glb" } }, { "sample_name": "062_PartNetMobility_small_appliances_small_appliances_1436", "benchmark_index": 62, "sample_id": "PartNetMobility/small_appliances/1436", "source_dataset": "PartNetMobility", "category": "small_appliances", "object_id": "small_appliances/1436", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 24, "texture_like_meshes": 9, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/062_PartNetMobility_small_appliances_small_appliances_1436/062_PartNetMobility_small_appliances_small_appliances_1436_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/062_PartNetMobility_small_appliances_small_appliances_1436@062_PartNetMobility_small_appliances_small_appliances_1436_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.5295761761871765, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 0, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 0, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 0, "status": "measured", "note": "" }, "joint_precision": { "value": 1.0, "status": "measured", "note": "" }, "joint_recall": { "value": 1.0, "status": "measured", "note": "" }, "joint_f1": { "value": 1.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 0.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.3437495308430837, "status": "measured", "note": "" }, "joint_limit_error": { "value": 53.354826601333905, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 1, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.47042382381282355, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 1.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 2, "gt_part_count": 2, "pred_joint_count": 1, "gt_joint_count": 1, "matches": [ { "type_mismatch": 0.0, "axis_error_deg": 0.0, "origin_error": 0.3437495308430837, "limit_error": 53.354826601333905, "accepted": true } ], "pred_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 52.95761761871765, "part_decomposition": 100.0, "kinematics": 100.0, "motion_consistency": 100.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 82.04593962400804, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/062_PartNetMobility_small_appliances_small_appliances_1436.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "PAct 在此样本上表现良好，与GT数据高度一致。\n\n1. **几何完整性与Part分解：** PAct成功地将物体分解为两个部分（中心圆盘和四个叶片），与GT parts=2一致。几何形状也与输入图像高度匹配。\n2. **关节数量/类型与运动可行性：** PAct识别出1个可动关节，与GT movable=1一致。考虑到这是一个风扇叶片结构，很可能是一个旋转关节，运动可行性高。\n3. **对2D分割的依赖：** PAct的输出与2D part mask高度一致，表明它能有效利用2D分割信息进行3D重建和part分解。\n4. **潜在问题：** 图像中风扇叶片似乎是连接在一起的，但2D mask将它们视为一个整体（红色部分）。如果GT将每个叶片视为一个单独的part，那么PAct的part分解可能不够细致。然而，根据提供的GT parts=2，PAct的分解是正确的。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "PAct 在此样本上表现良好，与GT数据高度一致。\n\n1. **几何完整性与Part分解：** PAct成功地将物体分解为两个部分（中心圆盘和四个叶片），与GT parts=2一致。几何形状也与输入图像高度匹配。\n2. **关节数量/类型与运动可行性：** PAct识别出1个可动关节，与GT movable=1一致。考虑到这是一个风扇叶片结构，很可能是一个旋转关节，运动可行性高。\n3. **对2D分割的依赖：** PAct的输出与2D part mask高度一致，表明它能有效利用2D分割信息进行3D重建和part分解。\n4. **潜在问题：** 图像中风扇叶片似乎是连接在一起的，但2D mask将它们视为一个整体（红色部分）。如果GT将每个叶片视为一个单独的part，那么PAct的part分解可能不够细致。然而，根据提供的GT parts=2，PAct的分解是正确的。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 419, "candidatesTokenCount": 230, "totalTokenCount": 649, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 161 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "dkEPat6mNtGHz7IPpMDLCA" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/062_PartNetMobility_small_appliances_small_appliances_1436/source_textured.glb", "gt_parts_glb": "viewer_assets/062_PartNetMobility_small_appliances_small_appliances_1436/gt_parts.glb", "pact_glb": "viewer_assets/062_PartNetMobility_small_appliances_small_appliances_1436/pact_parts.glb" } }, { "sample_name": "066_PartNetMobility_household_fixtures_household_fixtures_871", "benchmark_index": 66, "sample_id": "PartNetMobility/household_fixtures/871", "source_dataset": "PartNetMobility", "category": "household_fixtures", "object_id": "household_fixtures/871", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 5, "texture_like_meshes": 0, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/066_PartNetMobility_household_fixtures_household_fixtures_871/066_PartNetMobility_household_fixtures_household_fixtures_871_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/066_PartNetMobility_household_fixtures_household_fixtures_871@066_PartNetMobility_household_fixtures_household_fixtures_871_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.5278162570010345, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 0, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 0.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 1, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 0, "status": "measured", "note": "" }, "joint_precision": { "value": 0.0, "status": "measured", "note": "" }, "joint_recall": { "value": 0.0, "status": "measured", "note": "" }, "joint_f1": { "value": 0.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 90.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.0, "status": "measured", "note": "" }, "joint_limit_error": { "value": 0.0, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 0, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.4721837429989655, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 0.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 2, "gt_part_count": 2, "pred_joint_count": 1, "gt_joint_count": 1, "matches": [ { "type_mismatch": 0.0, "axis_error_deg": 90.0, "origin_error": 0.5901943181006489, "limit_error": 14.176057187728594, "accepted": false } ], "pred_tree": { "root_count": 2, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 0 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 52.781625700103454, "part_decomposition": 100.0, "kinematics": 0.0, "motion_consistency": 0.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 39.912888268437385, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/066_PartNetMobility_household_fixtures_household_fixtures_871.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "以下是对PAct在该样本上的简短测评和问题发现：\n\n1. **几何完整性与Part分解：** PAct生成的几何体与输入图像基本一致，part分解数量（2个）也与GT相符。然而，从2D mask看，红色部分（盖子上的小矩形）被识别为一个独立的part，这与实际物体（通常盖子是一个整体）不符，可能导致不自然的part分解。\n2. **关节数量/类型与运动可行性：** PAct的joint_f1为0.000，表明未能识别出任何关节，也未能识别出可动部件（GT movable=1）。这直接导致运动可行性为零，无法模拟物体（如垃圾桶盖）的铰链运动。\n3. **对2D分割的依赖：** PAct似乎过度依赖2D part mask的局部特征。2D mask中将盖子上的小矩形独立分割出来，PAct也将其视为一个独立part，但实际物体中这通常是盖子的一部分，而非独立可动部件。这暴露了模型对2D分割的强依赖性，且未能结合物体常识进行更合理的3D结构和运动推断。\n4. **整体性能：** PAct的weighted_score仅为39.91，且未能识别任何关节，表明其在理解物体铰链结构和运动能力方面表现不佳。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "以下是对PAct在该样本上的简短测评和问题发现：\n\n1. **几何完整性与Part分解：** PAct生成的几何体与输入图像基本一致，part分解数量（2个）也与GT相符。然而，从2D mask看，红色部分（盖子上的小矩形）被识别为一个独立的part，这与实际物体（通常盖子是一个整体）不符，可能导致不自然的part分解。\n2. **关节数量/类型与运动可行性：** PAct的joint_f1为0.000，表明未能识别出任何关节，也未能识别出可动部件（GT movable=1）。这直接导致运动可行性为零，无法模拟物体（如垃圾桶盖）的铰链运动。\n3. **对2D分割的依赖：** PAct似乎过度依赖2D part mask的局部特征。2D mask中将盖子上的小矩形独立分割出来，PAct也将其视为一个独立part，但实际物体中这通常是盖子的一部分，而非独立可动部件。这暴露了模型对2D分割的强依赖性，且未能结合物体常识进行更合理的3D结构和运动推断。\n4. **整体性能：** PAct的weighted_score仅为39.91，且未能识别任何关节，表明其在理解物体铰链结构和运动能力方面表现不佳。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 416, "candidatesTokenCount": 310, "totalTokenCount": 726, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 158 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "ekEPav6AHoDRz7IP5o_t6Ag" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/066_PartNetMobility_household_fixtures_household_fixtures_871/source_textured.glb", "gt_parts_glb": "viewer_assets/066_PartNetMobility_household_fixtures_household_fixtures_871/gt_parts.glb", "pact_glb": "viewer_assets/066_PartNetMobility_household_fixtures_household_fixtures_871/pact_parts.glb" } }, { "sample_name": "067_PartNetMobility_household_fixtures_household_fixtures_915", "benchmark_index": 67, "sample_id": "PartNetMobility/household_fixtures/915", "source_dataset": "PartNetMobility", "category": "household_fixtures", "object_id": "household_fixtures/915", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 11, "texture_like_meshes": 2, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/067_PartNetMobility_household_fixtures_household_fixtures_915/067_PartNetMobility_household_fixtures_household_fixtures_915_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/067_PartNetMobility_household_fixtures_household_fixtures_915@067_PartNetMobility_household_fixtures_household_fixtures_915_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.604920551819454, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 0, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 0, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 0, "status": "measured", "note": "" }, "joint_precision": { "value": 0.0, "status": "measured", "note": "" }, "joint_recall": { "value": 0.0, "status": "measured", "note": "" }, "joint_f1": { "value": 0.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 90.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.0, "status": "measured", "note": "" }, "joint_limit_error": { "value": 0.0, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 1, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.39507944818054597, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 0.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 2, "gt_part_count": 2, "pred_joint_count": 1, "gt_joint_count": 1, "matches": [ { "type_mismatch": 0.0, "axis_error_deg": 88.51979708892979, "origin_error": 0.3018219058890781, "limit_error": 359.63557525218357, "accepted": false } ], "pred_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 60.4920551819454, "part_decomposition": 100.0, "kinematics": 0.0, "motion_consistency": 0.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 41.13032450241243, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/067_PartNetMobility_household_fixtures_household_fixtures_915.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "以下是对PAct在该样本上的简短测评和问题发现：\n\n1. **几何完整性良好，Part分解准确：** PAct成功地从单视图输入和2D mask中重建了插座的几何形状，并且part分解（底座和开关）与GT parts=2一致。\n2. **关节数量/类型识别失败：** PAct的joint_f1为0.000，表明它未能识别出开关的可动性或正确的关节类型（GT movable=1）。\n3. **运动可行性缺失：** 由于未能识别关节，模型无法为开关生成可行的运动。\n4. **对2D分割的依赖性未完全体现：** 尽管2D part mask提供了明确的part边界，但模型未能将此信息转化为可动的关节，这可能表明其在从静态分割推断动态关节方面的局限性。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "以下是对PAct在该样本上的简短测评和问题发现：\n\n1. **几何完整性良好，Part分解准确：** PAct成功地从单视图输入和2D mask中重建了插座的几何形状，并且part分解（底座和开关）与GT parts=2一致。\n2. **关节数量/类型识别失败：** PAct的joint_f1为0.000，表明它未能识别出开关的可动性或正确的关节类型（GT movable=1）。\n3. **运动可行性缺失：** 由于未能识别关节，模型无法为开关生成可行的运动。\n4. **对2D分割的依赖性未完全体现：** 尽管2D part mask提供了明确的part边界，但模型未能将此信息转化为可动的关节，这可能表明其在从静态分割推断动态关节方面的局限性。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 416, "candidatesTokenCount": 194, "totalTokenCount": 610, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 158 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "f0EPasTyBPHoz7IP5NvmgQM" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/067_PartNetMobility_household_fixtures_household_fixtures_915/source_textured.glb", "gt_parts_glb": "viewer_assets/067_PartNetMobility_household_fixtures_household_fixtures_915/gt_parts.glb", "pact_glb": "viewer_assets/067_PartNetMobility_household_fixtures_household_fixtures_915/pact_parts.glb" } }, { "sample_name": "077_PartNetMobility_small_appliances_small_appliances_1450", "benchmark_index": 77, "sample_id": "PartNetMobility/small_appliances/1450", "source_dataset": "PartNetMobility", "category": "small_appliances", "object_id": "small_appliances/1450", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 14, "texture_like_meshes": 0, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/077_PartNetMobility_small_appliances_small_appliances_1450/077_PartNetMobility_small_appliances_small_appliances_1450_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/077_PartNetMobility_small_appliances_small_appliances_1450@077_PartNetMobility_small_appliances_small_appliances_1450_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.6677455169018812, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 0, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 0, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 0, "status": "measured", "note": "" }, "joint_precision": { "value": 0.0, "status": "measured", "note": "" }, "joint_recall": { "value": 0.0, "status": "measured", "note": "" }, "joint_f1": { "value": 0.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 90.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.0, "status": "measured", "note": "" }, "joint_limit_error": { "value": 0.0, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 1, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.3322544830981188, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 0.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 2, "gt_part_count": 2, "pred_joint_count": 1, "gt_joint_count": 1, "matches": [ { "type_mismatch": 0.0, "axis_error_deg": 90.0, "origin_error": 0.6575915599347089, "limit_error": 66.91012063819915, "accepted": false } ], "pred_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 66.77455169018812, "part_decomposition": 100.0, "kinematics": 0.0, "motion_consistency": 0.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 42.12229763529286, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/077_PartNetMobility_small_appliances_small_appliances_1450.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "以下是对PAct在给定样本上的简短测评和问题发现：\n\n1. **几何完整性与Part分解：** PAct生成的3D模型在几何上与输入图像高度一致，且Part分解与2D mask完全匹配，成功将风扇叶片和外壳分离。\n2. **关节数量/类型与运动可行性：** PAct未能识别出风扇叶片的可动性（joint_f1=0.000），表明其未能正确推断出旋转关节，导致运动可行性为零。\n3. **对2D分割的依赖：** PAct的Part分解完全依赖于2D part mask，这在几何上是准确的，但未能从2D信息中进一步推断出部件间的铰链关系和运动属性。\n4. **潜在问题：** 尽管part_count_mae=0，但未能识别出可动部件是主要缺陷，这限制了模型在生成可动3D物体方面的实用性。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "以下是对PAct在给定样本上的简短测评和问题发现：\n\n1. **几何完整性与Part分解：** PAct生成的3D模型在几何上与输入图像高度一致，且Part分解与2D mask完全匹配，成功将风扇叶片和外壳分离。\n2. **关节数量/类型与运动可行性：** PAct未能识别出风扇叶片的可动性（joint_f1=0.000），表明其未能正确推断出旋转关节，导致运动可行性为零。\n3. **对2D分割的依赖：** PAct的Part分解完全依赖于2D part mask，这在几何上是准确的，但未能从2D信息中进一步推断出部件间的铰链关系和运动属性。\n4. **潜在问题：** 尽管part_count_mae=0，但未能识别出可动部件是主要缺陷，这限制了模型在生成可动3D物体方面的实用性。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 419, "candidatesTokenCount": 219, "totalTokenCount": 638, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 161 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "gkEPar7oNaesz7IPxdi0mQo" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/077_PartNetMobility_small_appliances_small_appliances_1450/source_textured.glb", "gt_parts_glb": "viewer_assets/077_PartNetMobility_small_appliances_small_appliances_1450/gt_parts.glb", "pact_glb": "viewer_assets/077_PartNetMobility_small_appliances_small_appliances_1450/pact_parts.glb" } }, { "sample_name": "082_PartNetMobility_household_fixtures_household_fixtures_914", "benchmark_index": 82, "sample_id": "PartNetMobility/household_fixtures/914", "source_dataset": "PartNetMobility", "category": "household_fixtures", "object_id": "household_fixtures/914", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 216, "texture_like_meshes": 0, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/082_PartNetMobility_household_fixtures_household_fixtures_914/082_PartNetMobility_household_fixtures_household_fixtures_914_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/082_PartNetMobility_household_fixtures_household_fixtures_914@082_PartNetMobility_household_fixtures_household_fixtures_914_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.6380798192854266, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 0, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 0, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 0, "status": "measured", "note": "" }, "joint_precision": { "value": 0.0, "status": "measured", "note": "" }, "joint_recall": { "value": 0.0, "status": "measured", "note": "" }, "joint_f1": { "value": 0.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 90.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.0, "status": "measured", "note": "" }, "joint_limit_error": { "value": 0.0, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 1, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.36192018071457344, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 0.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 2, "gt_part_count": 2, "pred_joint_count": 1, "gt_joint_count": 1, "matches": [ { "type_mismatch": 0.0, "axis_error_deg": 89.756761982242, "origin_error": 0.43100563300486866, "limit_error": 359.7556539047208, "accepted": false } ], "pred_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 63.80798192854266, "part_decomposition": 100.0, "kinematics": 0.0, "motion_consistency": 0.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 41.65389188345411, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/082_PartNetMobility_household_fixtures_household_fixtures_914.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "以下是对 PAct 在该样本上的简短测评和问题发现：\n\n1. **几何完整性与 Part 分解：** PAct 似乎成功地根据 2D part mask 将物体分解为两个部分，与 GT parts 数量一致 (part_count_mae=0)。几何形状也基本保持完整。\n2. **关节数量/类型与运动可行性：** PAct 未能识别出任何关节 (joint_f1=0.000)，这表明它未能预测出物体的可动性，尽管 GT 显示有一个可动部件。这严重影响了运动可行性。\n3. **对 2D 分割的依赖：** PAct 严格遵循了 2D part mask 进行分割，但未能从几何结构中推断出潜在的铰链关系。这可能暴露出它对 2D 分割的过度依赖，而缺乏对 3D 结构和功能关系的理解。\n4. **PAct score 较低：** 41.65 的 weighted_score 相对较低，主要原因可能是未能识别出关节，导致 joint_f1 为 0。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "以下是对 PAct 在该样本上的简短测评和问题发现：\n\n1. **几何完整性与 Part 分解：** PAct 似乎成功地根据 2D part mask 将物体分解为两个部分，与 GT parts 数量一致 (part_count_mae=0)。几何形状也基本保持完整。\n2. **关节数量/类型与运动可行性：** PAct 未能识别出任何关节 (joint_f1=0.000)，这表明它未能预测出物体的可动性，尽管 GT 显示有一个可动部件。这严重影响了运动可行性。\n3. **对 2D 分割的依赖：** PAct 严格遵循了 2D part mask 进行分割，但未能从几何结构中推断出潜在的铰链关系。这可能暴露出它对 2D 分割的过度依赖，而缺乏对 3D 结构和功能关系的理解。\n4. **PAct score 较低：** 41.65 的 weighted_score 相对较低，主要原因可能是未能识别出关节，导致 joint_f1 为 0。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 416, "candidatesTokenCount": 248, "totalTokenCount": 664, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 158 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "hkEPas6DE9GHz7IPpMDLCA" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/082_PartNetMobility_household_fixtures_household_fixtures_914/source_textured.glb", "gt_parts_glb": "viewer_assets/082_PartNetMobility_household_fixtures_household_fixtures_914/gt_parts.glb", "pact_glb": "viewer_assets/082_PartNetMobility_household_fixtures_household_fixtures_914/pact_parts.glb" } }, { "sample_name": "089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543", "benchmark_index": 89, "sample_id": "PartNetMobility/major_appliances/PartNetMobility_12543", "source_dataset": "PartNetMobility", "category": "major_appliances", "object_id": "major_appliances/PartNetMobility_12543", "num_parts": 2, "movable_parts": 1, "texture_render": { "visual_meshes": 13, "texture_like_meshes": 6, "status": "textured_visual_render" }, "partcolor_debug_png": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_inputs/089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543/089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543_synthetic_partcolor_debug.png", "pact_object_json": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/pact_outputs/seed42_slatcfg7.0_sscfg7.0_sssteps25_slatsteps25_artioutmean_feature_regression_steps/exported_arti_objects/089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543@089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543_synthetic_processed.@ng/object.json", "pact_exploded_png": null, "metric_groups": { "pipeline_availability": { "inference_success": { "value": 1, "status": "measured", "note": "" }, "object_json_valid": { "value": 1, "status": "measured", "note": "" }, "glb_export_success": { "value": 0, "status": "not_provided", "note": "This smoke run exports articulated object JSON/PLY, not textured GLB." }, "part_asset_completeness": { "value": 1.0, "status": "measured", "note": "" }, "runtime_seconds": { "value": 14.901646375656128, "status": "measured", "note": "" }, "peak_gpu_memory_mb": { "value": null, "status": "not_provided", "note": "Not sampled in this run." }, "reproducible_seed_match": { "value": 1, "status": "measured", "note": "" } }, "geometry": { "chamfer_l2": { "value": null, "status": "not_provided", "note": "PLY-level Chamfer hook present; omitted in smoke run for speed." }, "fscore_tau_0_01": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "normal_consistency": { "value": null, "status": "not_provided", "note": "Requires dense surface sampling." }, "scale_consistency": { "value": 0.7448992539879123, "status": "measured", "note": "" }, "mesh_manifold_sanity": { "value": null, "status": "not_provided", "note": "Requires mesh repair/manifold pass." }, "texture_or_material_coverage": { "value": 0.0, "status": "not_provided", "note": "PAct smoke export is untextured PLY." } }, "part_decomposition": { "part_count_mae": { "value": 0, "status": "measured", "note": "" }, "part_count_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "part_segmentation_miou_if_masks_available": { "value": null, "status": "requires_gt", "note": "2D mask GT is conditioning input; not a predicted mask output." }, "semantic_part_accuracy": { "value": null, "status": "not_provided", "note": "PAct output names are not semantic labels in this export." }, "parent_child_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "tree_edit_distance": { "value": 0, "status": "measured", "note": "" }, "orphan_duplicate_part_rate": { "value": 0.0, "status": "measured", "note": "" } }, "kinematics": { "joint_count_mae": { "value": 0, "status": "measured", "note": "" }, "joint_precision": { "value": 0.0, "status": "measured", "note": "" }, "joint_recall": { "value": 0.0, "status": "measured", "note": "" }, "joint_f1": { "value": 0.0, "status": "measured", "note": "" }, "joint_type_accuracy": { "value": 1.0, "status": "measured", "note": "" }, "axis_angular_error_deg": { "value": 90.0, "status": "measured", "note": "" }, "origin_distance_error": { "value": 0.0, "status": "measured", "note": "" }, "joint_limit_error": { "value": 0.0, "status": "measured", "note": "" }, "kinematic_tree_validity": { "value": 1, "status": "measured", "note": "" } }, "motion_consistency": { "detachment_distance": { "value": null, "status": "requires_simulation", "note": "" }, "collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "self_collision_rate": { "value": null, "status": "requires_simulation", "note": "" }, "closed_state_assembly_error": { "value": 0.25510074601208765, "status": "measured", "note": "" }, "open_state_plausibility": { "value": 0.0, "status": "measured", "note": "" }, "trajectory_smoothness": { "value": null, "status": "requires_simulation", "note": "" } }, "ot_routing_diagnostics": { "ot_modules_active": { "value": 0, "status": "not_applicable", "note": "Raw PAct baseline; no OT module." }, "ot_gate_mean": { "value": null, "status": "not_applicable", "note": "" }, "ot_plan_marginal_error": { "value": null, "status": "not_applicable", "note": "" }, "ot_assignment_entropy": { "value": null, "status": "not_applicable", "note": "" }, "ot_edge_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_prior_beta": { "value": null, "status": "not_applicable", "note": "" }, "ot_mask_dropout_delta": { "value": null, "status": "not_applicable", "note": "" }, "ot_virtual_mass": { "value": null, "status": "not_applicable", "note": "" }, "part_patch_compactness": { "value": null, "status": "not_applicable", "note": "" } }, "robustness_breakdown": { "macro_score_by_source_dataset": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; see report.aggregate.by_source." }, "macro_score_by_category": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "macro_score_by_problem_tag": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "worst_10pct_score": { "value": null, "status": "not_applicable", "note": "Aggregate-only metric; sample count is 5." }, "failure_cluster_count": { "value": null, "status": "not_provided", "note": "Requires larger batch clustering." } }, "debug": { "pred_part_count": 2, "gt_part_count": 2, "pred_joint_count": 1, "gt_joint_count": 1, "matches": [ { "type_mismatch": 0.0, "axis_error_deg": 90.0, "origin_error": 0.5174011183546401, "limit_error": 83.3313706676152, "accepted": false } ], "pred_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 }, "ref_tree": { "root_count": 1, "duplicate_ids": 0, "dangling_parent_count": 0, "cycle_node_count": 0, "valid": 1 } } }, "group_scores": { "pipeline_availability": 100.0, "geometry": 74.48992539879123, "part_decomposition": 100.0, "kinematics": 0.0, "motion_consistency": 0.0, "ot_routing_diagnostics": 0.0 }, "weighted_score": 43.34051453665124, "vlm_card": "/data/250010098/PAct-Transporter/reports/20260513_pact_transporter_exploration/runs/pact_eval100_pm_easycase_probe_20260521/vlm_cards/089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543.png", "vlm": { "ok": true, "model": "gemini-2.5-flash", "text": "以下是对PAct在该样本上表现的简短测评和问题发现：\n\n1. **几何完整性与Part分解：** PAct成功地将物体分解为两个部分（与GT parts=2一致），并且几何形状大致完整，但手柄部分在PAct input中显得模糊且与主体融合，未被识别为独立部件。\n2. **关节数量/类型与运动可行性：** PAct的joint_f1为0.000，表明未能识别出任何关节，因此无法推断运动可行性。GT movable=1，说明存在一个可动部件，但PAct未能捕捉到。\n3. **对2D分割的依赖：** 2D part mask清晰地将顶部（红色）和底部（蓝色）分开，PAct的part分解与此2D分割高度一致。然而，2D mask未能区分手柄，这可能导致PAct也未能将其识别为独立部件或可动部件。\n4. **潜在问题：** PAct似乎过度依赖2D part mask进行部件分解，而未能从3D几何信息中推断出更精细的部件（如手柄）或关节信息，导致运动可行性评估失败。", "raw": { "candidates": [ { "content": { "parts": [ { "text": "以下是对PAct在该样本上表现的简短测评和问题发现：\n\n1. **几何完整性与Part分解：** PAct成功地将物体分解为两个部分（与GT parts=2一致），并且几何形状大致完整，但手柄部分在PAct input中显得模糊且与主体融合，未被识别为独立部件。\n2. **关节数量/类型与运动可行性：** PAct的joint_f1为0.000，表明未能识别出任何关节，因此无法推断运动可行性。GT movable=1，说明存在一个可动部件，但PAct未能捕捉到。\n3. **对2D分割的依赖：** 2D part mask清晰地将顶部（红色）和底部（蓝色）分开，PAct的part分解与此2D分割高度一致。然而，2D mask未能区分手柄，这可能导致PAct也未能将其识别为独立部件或可动部件。\n4. **潜在问题：** PAct似乎过度依赖2D part mask进行部件分解，而未能从3D几何信息中推断出更精细的部件（如手柄）或关节信息，导致运动可行性评估失败。" } ], "role": "model" }, "finishReason": "STOP", "index": 0 } ], "usageMetadata": { "promptTokenCount": 425, "candidatesTokenCount": 265, "totalTokenCount": 690, "promptTokensDetails": [ { "modality": "TEXT", "tokenCount": 167 }, { "modality": "IMAGE", "tokenCount": 258 } ], "serviceTier": "standard" }, "modelVersion": "gemini-2.5-flash", "responseId": "ikEPas7UJ-2Hz7IPvr-b-A4" } }, "interactive_3d": { "source_textured_glb": "viewer_assets/089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543/source_textured.glb", "gt_parts_glb": "viewer_assets/089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543/gt_parts.glb", "pact_glb": "viewer_assets/089_PartNetMobility_major_appliances_major_appliances_PartNetMobility_12543/pact_parts.glb" } } ], "aggregate": { "mean_weighted_score": 45.14436856104685, "by_source": { "PartNetMobility": 45.14436856104685 }, "mean_joint_f1": 0.125, "mean_part_count_mae": 0.125 } }