List your company

Spatial decision-making on FrozenLake size 12

64.72Success RateSpecFlow
20.67632.110543.54554.9795Jun 1, 2026

Evaluation results

MethodLinks
SpecFlowSetting=Flow(image) + AR(Text)2026.0664.72
Gemini-3-FlashSetting=Naive2026.0662.31
DiffThinkerSetting=Flow(image) Only2026.0660.41
GPT-5.1Setting=Naive2026.0658.74
MVoTSetting=Autoregressive2026.0658.62
Qwen3-VL-32BSetting=SFT2026.0647.36
Qwen3-VL-32BSetting=GRPO2026.0644.02
Qwen3-VL-8BSetting=SFT2026.0637.86
Qwen3-VL-8BSetting=GRPO2026.0634.56
Qwen3-VL-32BSetting=Naive2026.0633.06
Qwen3-VL-8BSetting=Naive2026.0622.37