Step Completion on CoSPlan Maze-E
46.1AccuracyGPT-4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-4oReasoning Strategy=Scene Graph2025.12 | 46.1 | |
| GPT-4oReasoning Strategy=Chain-of-Thought2025.12 | 45.6 | |
| Intern-VLMReasoning Strategy=Scene Graph2025.12 | 41.2 | |
| Intern-VLMReasoning Strategy=Chain-of-Thought2025.12 | 35.8 | |
| Qwen2 VL-8BReasoning Strategy=Scene Graph2025.12 | 28.3 | |
| Qwen2 VL-8BReasoning Strategy=Chain-of-Thought2025.12 | 27.9 | |
| Qwen2 VL-8BReasoning Strategy=Vanilla2025.12 | 26.5 | |
| CoG-VLMReasoning Strategy=Scene Graph2025.12 | 26.5 | |
| CoG-VLMReasoning Strategy=Chain-of-Thought2025.12 | 25.9 | |
| CoG-VLMReasoning Strategy=Vanilla2025.12 | 25.1 | |
| Janus-pro-7BReasoning Strategy=Scene Graph2025.12 | 21.7 | |
| Intern-VLMReasoning Strategy=Vanilla2025.12 | 21.6 | |
| Janus-pro-7BReasoning Strategy=Vanilla2025.12 | 20.4 | |
| Janus-pro-7BReasoning Strategy=Chain-of-Thought2025.12 | 20.2 | |
| Random2025.12 | 20 |