Robot Failure Analysis (MCQ) on RoboFAC (Real-world)
96FDGPT-4o
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4oFinetuned=false2026.04 | 96 | 43 | 52 | |
| KITE+Qwen2.5-7B+QLoRAFinetuned=true2026.04 | 89 | 58 | 77 | |
| KITE + Qwen2.5-VL-7BFinetuned=false2026.04 | 84 | 43 | 74 | |
| Qwen2.5-VL-7BFinetuned=false2026.04 | 83 | 38 | 72 | |
| RoboFAC-7BFinetuned=true2026.04 | 80 | 56 | 71 | |
| Gemini-2.0Finetuned=false2026.04 | 60 | 11 | 18 | |
| Qwen2.5-VL-3BFinetuned=false2026.04 | 4 | 3 | 7 |