List your company

Perception on V* (Pass@1)

90.2Pass@1SFT + GRPO
61.91269.25676.683.944May 27, 2026

Evaluation results

MethodLinks
SFT + GRPOBackbone=Qwen3-VL-Thinking, Model Size=4B2026.0590.2
BaseBackbone=Qwen3-VL-Thinking, Model Size=32B2026.0589.1
SFT + AXPOBackbone=Qwen3-VL-Thinking, Model Size=4B2026.0588.9
SFT + AXPOBackbone=Qwen3-VL-Thinking, Model Size=8B2026.0587.8
SFT + GRPOBackbone=Qwen3-VL-Thinking, Model Size=8B2026.0587.7
SFTBackbone=Qwen3-VL-Thinking, Model Size=4B2026.0587
GRPOBackbone=Qwen3-VL-Thinking, Model Size=8B2026.0585.7
SFTBackbone=Qwen3-VL-Thinking, Model Size=8B2026.0584.8
GRPOBackbone=Qwen3-VL-Thinking, Model Size=4B2026.0582.7
SFT + GRPOBackbone=Qwen3-VL-Thinking, Model Size=2B2026.0581.7
SFT + AXPOBackbone=Qwen3-VL-Thinking, Model Size=2B2026.0581.3
BaseBackbone=Qwen3-VL-Thinking, Model Size=4B2026.0580.6
BaseBackbone=Qwen3-VL-Thinking, Model Size=8B2026.0577.7
SFTBackbone=Qwen3-VL-Thinking, Model Size=2B2026.0575.9
GRPOBackbone=Qwen3-VL-Thinking, Model Size=2B2026.0567.1
BaseBackbone=Qwen3-VL-Thinking, Model Size=2B2026.0563