Embodied Task Success on EmbodiedBench EB-Navigation 1.0 (test)
61.33Average Success RateEmbodiSkill
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EmbodiSkillExecutor=Qwen3-VL-32B-Instruct, Skill Evolution Model=GPT-5.22026.05 | 61.33 | 65 | 68 | 73 | 67 | 33 | |
| EmbodiSkillExecutor=Qwen3-VL-32B-Instruct, Skill Evolution Model=Gemini-3-flash2026.05 | 61.33 | 67 | 70 | 70 | 68 | 32 | |
| GPT-5.2Executor=Direct (Closed-source)2026.05 | 57.33 | 63 | 72 | 68 | 65 | 18 | |
| Gemini-3-flashExecutor=Direct (Closed-source)2026.05 | 56 | 62 | 60 | 55 | 47 | 57 | |
| Mem0Executor=Qwen3-VL-32B-Instruct, Memory Strategy=Mem02026.05 | 52 | 62 | 60 | 65 | 67 | 7 | |
| LangMemExecutor=Qwen3-VL-32B-Instruct, Memory Strategy=LangMem2026.05 | 52 | 62 | 60 | 65 | 67 | 7 | |
| EmbodiSkillExecutor=Qwen3-VL-8B-Instruct, Skill Evolution Model=GPT-5.22026.05 | 50.33 | 68 | 65 | 62 | 57 | 0 | |
| G-MemoryExecutor=Qwen3-VL-32B-Instruct, Memory Strategy=G-Memory2026.05 | 50.33 | 52 | 53 | 58 | 65 | 23 | |
| EmbodiSkillExecutor=Qwen3-VL-8B-Instruct, Skill Evolution Model=Gemini-3-flash2026.05 | 49 | 60 | 63 | 67 | 52 | 3 | |
| LangMemExecutor=Qwen3-VL-8B-Instruct, Memory Strategy=LangMem2026.05 | 47 | 60 | 57 | 62 | 57 | 0 | |
| Mem0Executor=Qwen3-VL-8B-Instruct, Memory Strategy=Mem02026.05 | 46.33 | 58 | 57 | 62 | 55 | 0 | |
| No memoryExecutor=Qwen3-VL-32B-Instruct, Memory Strategy=None2026.05 | 46.33 | 60 | 53 | 60 | 53 | 5 | |
| No memoryExecutor=Qwen3-VL-8B-Instruct, Memory Strategy=None2026.05 | 45 | 58 | 53 | 62 | 50 | 2 | |
| G-MemoryExecutor=Qwen3-VL-8B-Instruct, Memory Strategy=G-Memory2026.05 | 37.33 | 62 | 43 | 43 | 38 | 0 |