Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Sokoban on Sokoban (test)
Loading...
90.6
Success Rate
HiMAC
11.04
31.695
52.35
73.005
Mar 1, 2026
Success Rate
Score
Updated 1mo ago
Evaluation Results
Method
Method
Links
Success Rate
Score
HiMAC
Backbone=Qwen3-VL-8B
2026.03
90.6
7.32
HiMAC
Backbone=Qwen2.5-VL-7B
2026.03
87.5
6.7
GRPO
Backbone=Qwen2.5-VL-7B
2026.03
83.9
5.39
HiMAC
Backbone=Qwen2.5-VL-3B
2026.03
83.8
5.31
GRPO
Backbone=Qwen3-VL-8B
2026.03
82.9
5.24
GiGPO
Backbone=Qwen2.5-VL-7B
2026.03
82.8
5.27
GiGPO
Backbone=Qwen2.5-VL-3B
2026.03
81.1
4.86
GiGPO
Backbone=Qwen3-VL-8B
2026.03
78.1
4.67
GRPO
Backbone=Qwen2.5-VL-3B
2026.03
76.2
4.59
Prompt
Backbone=Qwen2.5-VL-7B
2026.03
25
-0.26
Prompt
Backbone=Qwen3-VL-8B
2026.03
21.9
-0.95
Prompt
Backbone=Qwen2.5-VL-3B
2026.03
14.1
-0.88
Feedback
Search any
task
Search any
task