Robotic Failure Analysis VQA on ViFailback-Bench Lite
54.64AccuracyGemini-2.5-Pro
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini-2.5-ProModel Category=General Closed-Source Models2025.12 | 54.64 | |
| Qwen2.5-VL-72B-InstructModel Category=General Open-Source Models2025.12 | 50.61 | |
| RoboBrain2.0-32BModel Category=Embodied Models2025.12 | 49.92 | |
| GPT-4oModel Category=General Closed-Source Models2025.12 | 48.21 | |
| Qwen3-VL-32B-InstructModel Category=General Open-Source Models2025.12 | 47.79 | |
| Qwen2.5-VL-32B-InstructModel Category=General Open-Source Models2025.12 | 46.3 | |
| InternVL3-78BModel Category=General Open-Source Models2025.12 | 42.81 | |
| Qwen2.5-VL-7B-InstructModel Category=General Open-Source Models2025.12 | 42.41 | |
| Qwen3-VL-4B-InstructModel Category=General Open-Source Models2025.12 | 41.11 | |
| RoboBrain2.0-7BModel Category=Embodied Models2025.12 | 40.62 | |
| RoboBrain2.0-3BModel Category=Embodied Models2025.12 | 40.39 | |
| Qwen3-VL-8B-InstructModel Category=General Open-Source Models2025.12 | 38.33 | |
| Qwen2.5-VL-3B-InstructModel Category=General Open-Source Models2025.12 | 38.1 | |
| Cosmos-Reason1-7BModel Category=Embodied Models2025.12 | 38.06 | |
| InternVL3-8BModel Category=General Open-Source Models2025.12 | 36.48 | |
| Qwen3-VL-2B-InstructModel Category=General Open-Source Models2025.12 | 35.16 |