List your company

Visual Reasoning on Ball Tracking

77.33AccuracyGemini-1.5-Flash
27.4140.3753.3366.29Jun 11, 2026

Evaluation results

MethodLinks
Gemini-1.5-FlashModel Category=Frontier Models2026.0677.33
Ours (Stage 2: optimize Lund→gen)Optimization Stage=Stage 2 (Flow-GRPO), Model Category=Our Models2026.0672.67
Ours (Stage 1: optimize Lgen→und)Optimization Stage=Stage 1 (Reflective SFT), Model Category=Our Models2026.0670.33
Seed-1.5-27BModel Category=Frontier Models2026.0658.67
ThinkMorphModel Category=Open-Source Models, Interleave Thinking=true2026.0648
Gemini-1.5-Flash-8BModel Category=Frontier Models2026.0640
ThinkMorphModel Category=Open-Source Models, Interleave Thinking=false2026.0637.33
Qwen2.5-7BModel Category=Open-Source Models2026.0632
Gemma-2-9BModel Category=Open-Source Models2026.0632
Bagel-7B-MoTModel Category=Open-Source Models2026.0629.33