File size: 3,402 Bytes
f89c6d8 | 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 | {
"schema_version": "mindview_evaluation_protocol_v1",
"split": "test",
"index": "data/test.jsonl",
"index_sha256": "de79a2a551d31117e4ad87b4cfaee7b47c032749743bba24ee162c3c9dc53383",
"examples": 1500,
"scenes": 513,
"tasks": {
"translation": "Translation",
"rotation": "Rotation",
"composition": "Composition",
"obj_loc": "Relational Object Grounding",
"loc-re-facing": "Target-Facing Reorientation",
"anchor_guided_relative_repositioning": "Object-Centric Repositioning",
"view_localization": "Multi-Cue Viewpoint Inference"
},
"metrics": [
"DINO-I",
"Semantic",
"Spatial",
"Quality",
"VGGT-RScore",
"VGGT-TScore"
],
"metric_definitions": [
{
"name": "dino_i",
"label": "DINO-I (x100)",
"source": "clip_dino",
"field": "dino_i",
"scale": 100.0,
"raw_range": [
-1.0,
1.0
],
"direction": "higher"
},
{
"name": "semantic",
"label": "Semantic",
"source": "qwen_vl",
"field": "semantic_score_100",
"scale": 1.0,
"raw_range": [
0.0,
100.0
],
"direction": "higher"
},
{
"name": "spatial",
"label": "Spatial",
"source": "qwen_vl",
"field": "spatial_score_100",
"scale": 1.0,
"raw_range": [
0.0,
100.0
],
"direction": "higher"
},
{
"name": "quality",
"label": "Visual Quality",
"source": "qwen_vl",
"field": "visual_quality_score_100",
"scale": 1.0,
"raw_range": [
0.0,
100.0
],
"direction": "higher"
},
{
"name": "vggt_rscore",
"label": "VGGT-RScore",
"source": "vggt",
"field": "rscore_0_100",
"scale": 1.0,
"raw_range": [
0.0,
100.0
],
"direction": "higher"
},
{
"name": "vggt_tscore",
"label": "VGGT-TScore",
"source": "vggt",
"field": "tscore_0_100",
"scale": 1.0,
"raw_range": [
0.0,
100.0
],
"direction": "higher"
}
],
"per_metric_aggregation": "Mean within each task, then equal-weight mean over the seven tasks.",
"overall_aggregation": "Unweighted arithmetic mean of the six task-macro primary metric scores on their reported scales.",
"auxiliary_metrics_excluded_from_overall": [
"CLIP-I"
],
"generation_inputs": "Use only the ordered reference RGB inputs and instruction prescribed by each test record.",
"scoring_index": "evaluation/references.jsonl",
"scoring_reference_count": 10,
"scoring_reference_order": "Frozen ordered references; do not replace with the generator's shorter input list.",
"held_out_target": "Target RGB and depth are evaluation assets; never provide them as generator inputs.",
"depth": {
"storage": "Original unsigned 16-bit PNG sensor depth",
"unit": "millimeter",
"invalid_values": [
0,
65535
],
"runtime_meter_conversion_divisor": 1000
},
"context_scaling": "Separate optional experiment; do not pool its configurations into the official score.",
"runtime_integration_status": "pending_validation_of_official_split_entrypoint",
"runtime_caveat": "Historical report configurations are tied to a combined 4500-record index and cannot be assumed compatible by substituting this test index alone."
}
|