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

General Multimodal Reasoning on NaturalBench

78.62General Score

Qwen2.5-VL-7B + SRPO

Updated 23d ago

Evaluation Results

Method	Links
Qwen2.5-VL-7B + SRPO 2026.05		78.62
VL-Rethinker-7B 2026.05		78.17
PAPO-D-7B 2026.05		78.11
ThinkLite-VL-7B 2026.05		78.09
Qwen2.5-VL-7B + DAPO 2026.05		78.05
Vision-Matters-7B 2026.05		78.01
Perception-R1-7B 2026.05		77.96
VPPO-7B 2026.05		77.94
PAPO-G-7B 2026.05		77.92
Qwen2.5-VL-7B + GRPO 2026.05		77.92
PRCO-7B 2026.05		77.83
Qwen2.5-VL-7B 2026.05		77.67
Vision-SR1-7B 2026.05		77.65
Qwen2.5-VL-3B + SRPO 2026.05		74.83
PAPO-G-3B 2026.05		74.76
Qwen2.5-VL-3B + DAPO 2026.05		74.61
PAPO-D-3B 2026.05		74.45
PRCO-3B 2026.05		74.31
Qwen2.5-VL-3B + GRPO 2026.05		74.11
Vision-SR1-3B 2026.05		73.83
Qwen2.5-VL-3B 2026.05		72.27