List your company

Spatial Decision-Making on FrozenLake size 8

81.74Success RateGemini-3-Flash
35.470447.482759.49571.5073Jun 1, 2026

Evaluation results

MethodLinks
Gemini-3-FlashSetting=Naive2026.0681.74
SpecFlowSetting=Flow(image) + AR(Text)2026.0681.73
DiffThinkerSetting=Flow(image) Only2026.0679.83
GPT-5.1Setting=Naive2026.0679.68
MVoTSetting=Autoregressive2026.0678.94
Qwen3-VL-32BSetting=SFT2026.0662.91
Qwen3-VL-32BSetting=GRPO2026.0659.47
Qwen3-VL-8BSetting=SFT2026.0653.72
Qwen3-VL-8BSetting=GRPO2026.0650.21
Qwen3-VL-32BSetting=Naive2026.0648.12
Qwen3-VL-8BSetting=Naive2026.0637.25