Personalized Embodied Navigation on PinNED (Distractor)
60Success Rate (SR)POLAR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| POLARBackbone=Qwen2.5-VL-8B-Instruct2026.05 | 60 | 26 | |
| POLARBackbone=GPT-52026.05 | 37.5 | 14.73 | |
| no-priorBackbone=Qwen2.5-VL-8B-Instruct2026.05 | 30 | 27.77 | |
| POLARBackbone=GPT-4o-mini2026.05 | 30 | 3.36 | |
| no-priorBackbone=GPT-52026.05 | 25 | 14.73 | |
| raw-interactionBackbone=GPT-52026.05 | 25 | 11.2 | |
| raw-interactionBackbone=Gemini2026.05 | 25 | 3.24 | |
| POLARBackbone=Gemini2026.05 | 25 | 12.95 | |
| POLARBackbone=Qwen3-VL-8B-Instruct2026.05 | 20 | 18 | |
| raw-interactionBackbone=Qwen2.5-VL-8B-Instruct2026.05 | 20 | 9.06 | |
| no-priorBackbone=GPT-4o-mini2026.05 | 20 | 2.24 | |
| raw-interactionBackbone=GPT-4o-mini2026.05 | 20 | 2.24 | |
| no-priorBackbone=Gemini2026.05 | 10 | 7.7 | |
| no-priorBackbone=Qwen3-VL-8B-Instruct2026.05 | 0 | 0 | |
| raw-interactionBackbone=Qwen3-VL-8B-Instruct2026.05 | 0 | 0 |