arxiv
Qwen3-VL GRPO
2 benchmarks
DocVQA
#4 of 15
95.9%
MMMU
#14 of 24
69.7%
benchmark.space canonical public ledger of AI capability