Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Compositional Reasoning on Lego Puzzles
Loading...
57.7
Accuracy
GPT-4o-0513
25.772
34.061
42.35
50.639
Nov 10, 2025
Accuracy
Updated 18d ago
Evaluation Results
Method
Method
Links
Accuracy
GPT-4o-0513
2025.11
57.7
Claude-3.5-Sonnet-0620
2025.11
53.6
SpatialThinker-7B (Ours)
Training Data=SpatialT...
2025.11
37.7
Qwen2.5-VL-7B + SFT
Training Data=SpatialT...
2025.11
36.6
Qwen2.5-VL-7B
2025.11
35.8
Qwen2.5-VL-3B + SFT
Training Data=SpatialT...
2025.11
34.7
SpatialThinker-3B (Ours)
Training Data=SpatialT...
2025.11
33.9
VLAA-Thinker-7B
2025.11
33.4
SpaceOm
2025.11
32
SpaceThinker
2025.11
31.5
Qwen2.5-VL-3B
2025.11
29.9
Qwen2.5-VL-7B + Vanilla GRPO
Training Data=SpatialT...
2025.11
29.7
Qwen2.5-VL-3B + Vanilla GRPO
Training Data=SpatialT...
2025.11
27
Feedback
Search any
task
Search any
task