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

Vision-centric Reasoning on CV Bench

83.8Accuracy

Long Grounded Thoughts (SFT + GRPO)

72.328875.306978.28581.2631Nov 7, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.11
83.8
2025.11
82.3
2025.11
82.28
2025.11
81.8
2025.11
81.51
2025.11
80.6
2025.11
76
2025.11
75.4
2025.11
75.3
2025.11
74.52
2025.11
72.95
2025.11
72.77