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

Fine-grained Visual Reasoning on V-Star, HR-Bench-4K, and HR-Bench-8K

85.9Average Score

P2R-8B

Updated 23d ago

Evaluation Results

Method	Links
P2R-8B 2026.07		85.9
P2R-4B 2026.07		85.2
Qwen3-VL-Instruct-32B 2026.07		79.7
DeepEyes-7B 2026.07		79.3
P2R-2B 2026.07		78.1
Thyme-7B 2026.07		77.1
ZoomEye-7B 2026.07		76.5
Qwen3-VL-Instruct-8B 2026.07		76.2
Qwen3-VL-Instruct-4B 2026.07		74.2
PixelReasoner-7B 2026.07		73.5
Qwen3-VL-Instruct-2B 2026.07		70
GPT-4o 2026.07		60