Temporal Spatial Reasoning on TempCompass
76.8Average AccuracyG2F-RAG
Evaluation Results
| Method | Links | |
|---|---|---|
| G2F-RAGSize=14B, Base Model=InternVL3.52026.04 | 76.8 | |
| G2F-RAGSize=8B, Base Model=InternVL3.52026.04 | 75.5 | |
| G2F-RAGSize=7B, Base Model=Qwen2.5-VL2026.04 | 72 | |
| InternVL3.5Size=14B2026.04 | 71.2 | |
| InternVL3.5Size=8B2026.04 | 71 | |
| InternVL3Size=8B2026.04 | 70.4 | |
| GPT-4o (2024-05-13)Size=-2026.04 | 69.5 | |
| Qwen2.5-VLSize=7B2026.04 | 69.2 | |
| G2F-RAGSize=7B, Base Model=LLaVA-Video2026.04 | 69 | |
| G2F-RAGSize=4B, Base Model=InternVL3.52026.04 | 68.7 | |
| InternVL2.5Size=8B2026.04 | 68.7 | |
| Gemini1.5 ProSize=-2026.04 | 68.4 | |
| LLaVA-VideoSize=7B2026.04 | 65.5 | |
| InternVL3.5Size=4B2026.04 | 65.4 | |
| Qwen2.5-VLSize=3B2026.04 | 64.4 | |
| MiniCPM-V 2.6Size=8B2026.04 | 59.6 | |
| VILA-1.5Size=8B2026.04 | 58.8 |