Object Goal Navigation on CHORES-S 1.0 (test)
81Success Rate (SR)HiRO-Nav
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HiRO-NavTraining Recipe=H-SFT+RLI&II, Reasoning Strategy=Ours (Hybrid), ASM=✓, Dataset=HRD2026.04 | 81 | 57.2 | 2.7 | 13.5 | |
| HiRO-NavTraining Recipe=H-SFT+RLI&II, Reasoning Strategy=Thinking-Every-K-Steps, ASM=✓, Dataset=HRD2026.04 | 78.5 | 46.7 | 3.5 | 17.3 | |
| Qwen2.5VL-3BTraining Recipe=SFT+RLI, Reasoning Strategy=No-Thinking, ASM=✓, Dataset=NRD2026.04 | 70.5 | 52.5 | 8.1 | 3.7 | |
| HiRO-NavTraining Recipe=H-SFT+RLI&II, Reasoning Strategy=No-Thinking, ASM=✓, Dataset=HRD2026.04 | 70 | 48.8 | 8.2 | 3.6 | |
| Qwen2.5VL-3BTraining Recipe=H-SFT, Reasoning Strategy=Thinking-Every-K-Steps, ASM=✓, Dataset=HRD2026.04 | 60 | 40.3 | 4.8 | 17.6 | |
| HiRO-NavTraining Recipe=H-SFT, Reasoning Strategy=Ours (Hybrid), ASM=✓, Dataset=HRD2026.04 | 59.5 | 48.9 | 2.3 | 9.2 | |
| Qwen2.5VL-3BTraining Recipe=SFT, Reasoning Strategy=No-Thinking, ASM=✓, Dataset=NRD2026.04 | 50 | 41.2 | 1.2 | 3.8 | |
| Qwen2.5VL-3BTraining Recipe=H-SFT, Reasoning Strategy=No-Thinking, ASM=✓, Dataset=HRD2026.04 | 49.5 | 39.1 | 1.2 | 3.8 | |
| NaVILATraining Recipe=SFT+RLI, Reasoning Strategy=No-Thinking, ASM=×, Dataset=NRD2026.04 | 44 | 27.1 | 1.5 | 3.8 | |
| Qwen2.5VL-3BTraining Recipe=SFT, Reasoning Strategy=No-Thinking, ASM=×, Dataset=NRD2026.04 | 36.5 | 28.8 | 1.5 | 3.8 | |
| HiRO-NavTraining Recipe=H-SFT+RLI&II, Reasoning Strategy=Dense-Thinking, ASM=✓, Dataset=HRD2026.04 | 34.5 | 15.6 | 3.1 | 70.4 | |
| NaVILATraining Recipe=SFT, Reasoning Strategy=No-Thinking, ASM=×, Dataset=NRD2026.04 | 30.5 | 12.1 | 1.8 | 3.7 | |
| Gemini2.5-ProTraining Recipe=TF, Reasoning Strategy=Dense-Thinking, ASM=✓, Dataset=-2026.04 | 27.5 | 16.2 | 2.3 | 804.8 | |
| o3Training Recipe=TF, Reasoning Strategy=Dense-Thinking, ASM=✓, Dataset=-2026.04 | 27 | 12.3 | 2.4 | 539.4 | |
| Qwen2.5VL-3BTraining Recipe=H-SFT, Reasoning Strategy=Dense-Thinking, ASM=✓, Dataset=HRD2026.04 | 21 | 8.2 | 3.6 | 72.1 | |
| GPT-4oTraining Recipe=TF, Reasoning Strategy=No-Thinking, ASM=✓, Dataset=-2026.04 | 8 | 4.4 | 5.6 | 3.1 |