Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

SPAR-Bench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Spatial ReasoningSPAR-Bench
Overall Score67.3
45
Spatial Reasoning (Multi-Image)SPAR-Bench
Accuracy67.3
28
Spatial Relationship ReasoningSPAR-Bench
Accuracy (Avg)59.9
26
Spatial ReasoningSPAR-Bench full
Average Score68.35
23
Single-image spatial reasoningSPAR-Bench SI
Low Score54.3
15
Multi-image Spatial ReasoningSPAR-Bench-MV (test)
Score (Low Difficulty)43.7
15
Spatial ReasoningSPAR-Bench tiny
Medium Difficulty Score72.32
12
Spatial ReasoningSPAR-Bench SI, MV 91
Accuracy63.3
11
High-level spatial reasoningSPAR-Bench high-level tasks
High Average Score51.28
8
Showing 9 of 9 rows