2D Navigation on FrozenLake OOD
98Success RateGPT-5-Thinking-Tool
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-5-Thinking-ToolAccess level (Closed-source/Open-source)=Closed-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Chain-of-thought (CoT) / Thinking mode=Thinking-Tool2025.09 | 98 | 77 | |
| Uni-PlanModel scale (Parameters)=14B-A7B, Planning planner (Open-loop/Closed-loop)=Open-loop2025.09 | 95 | 70 | |
| VLPAccess level (Closed-source/Open-source)=Open-Source, Planning planner (Open-loop/Closed-loop)=Open-loop2025.09 | 94 | 56 | |
| Qwen2.5-VL-32B-Ins-CoTAccess level (Closed-source/Open-source)=Open-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Model scale (Parameters)=32B, Chain-of-thought (CoT) / Thinking mode=True2025.09 | 49 | 44 | |
| GPT-5-ThinkingAccess level (Closed-source/Open-source)=Closed-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Chain-of-thought (CoT) / Thinking mode=True2025.09 | 44 | 59 | |
| Qwen2.5-VL-32B-InsAccess level (Closed-source/Open-source)=Open-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Model scale (Parameters)=32B, Chain-of-thought (CoT) / Thinking mode=False2025.09 | 43 | 25 | |
| BAGEL-VLM-CoT (baseline, 7B)Access level (Closed-source/Open-source)=Open-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Model scale (Parameters)=7B, Chain-of-thought (CoT) / Thinking mode=True2025.09 | 43 | 41 | |
| BAGEL-VLM (baseline, 7B)Access level (Closed-source/Open-source)=Open-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Model scale (Parameters)=7B, Chain-of-thought (CoT) / Thinking mode=False2025.09 | 38 | 21 | |
| Qwen2.5-VL-7B-Ins-CoTAccess level (Closed-source/Open-source)=Open-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Model scale (Parameters)=7B, Chain-of-thought (CoT) / Thinking mode=True2025.09 | 37 | 26 | |
| Qwen2.5-VL-7B-InsAccess level (Closed-source/Open-source)=Open-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Model scale (Parameters)=7B, Chain-of-thought (CoT) / Thinking mode=False2025.09 | 33 | 14 | |
| GPT-5Access level (Closed-source/Open-source)=Closed-Source, Planning planner (Open-loop/Closed-loop)=Open-loop, Chain-of-thought (CoT) / Thinking mode=False2025.09 | 8 | 12 |