Embodied AI Task Execution on EB-ALFRED online unsupervised setting
61Success Rate (Avg)ELITE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ELITEBackbone=Qwen2.5-VL-72B-Ins2026.03 | 61 | 60 | 62 | |
| Qwen2.5-VL-72B-InsModel Type=Open-Source MLLM2026.03 | 52 | 55 | 49 | |
| ELITEBackbone=InternVL3-78B2026.03 | 44 | 40 | 48 | |
| InternVL2.5-78BModel Type=Open-Source MLLM2026.03 | 40 | 38 | 42 | |
| ESCABackbone=Qwen2.5-VL-72B-Ins2026.03 | 38 | 46 | 30 | |
| InternVL3-78BModel Type=Open-Source MLLM2026.03 | 37 | 38 | 36 | |
| Qwen2-VL-72B-InsModel Type=Open-Source MLLM2026.03 | 35 | 40 | 30 | |
| gemma-3-27b-itModel Type=Open-Source MLLM2026.03 | 34 | 42 | 26 | |
| Ovis2-34BModel Type=Open-Source MLLM2026.03 | 29 | 34 | 24 | |
| Llama-3.2-90B-Vision-InsModel Type=Open-Source MLLM2026.03 | 27 | 38 | 16 |