Visual Question Answering on GQA (Exact Match)
63.3Exact MatchARGUSVLM
Evaluation Results
| Method | Links | |
|---|---|---|
| ARGUSVLM#Parameters=2B2026.03 | 63.3 | |
| VanillaBackbone=LLaVA-1.5-7B, Retained Tokens=5762026.06 | 61.9 | |
| Qwen2.5-VL*#Parameters=3B2026.03 | 60.2 | |
| MiniGemini#Parameters=2.3B2026.03 | 59.9 | |
| Qwen3-VL*#Parameters=2B2026.03 | 59.4 | |
| MobileVLM-V2#Parameters=1.7B2026.03 | 59.3 | |
| DeepSeekVL#Parameters=2B2026.03 | 59.3 | |
| InternVL3*#Parameters=2B2026.03 | 58.9 | |
| DCP-PruneBackbone=LLaVA-1.5-7B, Retained Tokens=642026.06 | 58.5 | |
| SCOPEBackbone=LLaVA-1.5-7B, Retained Tokens=642026.06 | 58.3 | |
| ARGUSVLM#Parameters=500M2026.03 | 58.2 | |
| PRUNESIDBackbone=LLaVA-1.5-7B, Retained Tokens=642026.06 | 57.1 | |
| DCP-PruneBackbone=LLaVA-1.5-7B, Retained Tokens=322026.06 | 57 | |
| SCOPEBackbone=LLaVA-1.5-7B, Retained Tokens=322026.06 | 56.4 | |
| MobileVLM#Parameters=1.7B2026.03 | 56.1 | |
| FlowCutBackbone=LLaVA-1.5-7B, Retained Tokens=642026.06 | 55.6 | |
| ARGUSVLM#Parameters=256M2026.03 | 55.3 | |
| HoloVBackbone=LLaVA-1.5-7B, Retained Tokens=642026.06 | 55.3 | |
| VisionZipBackbone=LLaVA-1.5-7B, Retained Tokens=642026.06 | 55.1 | |
| DCP-PruneBackbone=LLaVA-1.5-7B, Retained Tokens=162026.06 | 55.1 | |
| PRUNESIDBackbone=LLaVA-1.5-7B, Retained Tokens=322026.06 | 54.8 | |
| SCOPEBackbone=LLaVA-1.5-7B, Retained Tokens=162026.06 | 53.4 | |
| HoloVBackbone=LLaVA-1.5-7B, Retained Tokens=322026.06 | 52.8 | |
| VisionZipBackbone=LLaVA-1.5-7B, Retained Tokens=322026.06 | 51.9 | |
| FlowCutBackbone=LLaVA-1.5-7B, Retained Tokens=322026.06 | 51.9 | |
| PRUNESIDBackbone=LLaVA-1.5-7B, Retained Tokens=162026.06 | 50.3 | |
| SmolVLM*#Parameters=2B2026.03 | 49.2 | |
| HoloVBackbone=LLaVA-1.5-7B, Retained Tokens=162026.06 | 48.9 | |
| FlowCutBackbone=LLaVA-1.5-7B, Retained Tokens=162026.06 | 48.2 | |
| VisionZipBackbone=LLaVA-1.5-7B, Retained Tokens=162026.06 | 47 | |
| SmolVLM*#Parameters=500M2026.03 | 45.3 | |
| SmolVLM*#Parameters=256M2026.03 | 41.5 |