Visual Question Answering on TextVQA (test val)
58.2AccuracyPromptCD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PromptCD2026.02 | 58.2 | 1.34 | |
| VanillaVenue=NeurIPS’23, Location=—, Tokens=576, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 58.2 | — | |
| SVD-PruneVenue=CVPR’25, Location=Encoder, Tokens=192, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 57.24 | — | |
| PyramidDropVenue=CVPR’25, Location=LLM, Tokens=128, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 56.6 | — | |
| PyramidDropVenue=AAAI’25, Location=LLM, Tokens=192, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 56.5 | — | |
| SVD-PruneVenue=—, Location=Encoder, Tokens=128, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 56.14 | — | |
| ViCropvariant=grad-att2026.02 | 56.06 | 0.89 | |
| VisionZipVenue=CVPR’25, Location=Encoder, Tokens=128, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 55.8 | — | |
| ViCropvariant=rel-att2026.02 | 55.17 | 1.16 | |
| SVD-PruneVenue=—, Location=Encoder, Tokens=64, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 55.14 | — | |
| SVD-PruneVenue=—, Location=Encoder, Tokens=32, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 54.81 | — | |
| SVD-PruneVenue=—, Location=Encoder, Tokens=16, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 54.03 | — | |
| SparseVLMVenue=ICLR’25, Location=LLM, Tokens=64, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 53.4 | — | |
| VisionZipVenue=CVPR’25, Location=Encoder, Tokens=32, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 53.1 | — | |
| FastVVenue=ECCV’24, Location=LLM, Tokens=192, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 52.5 | — | |
| ToMeVenue=ICLR’23, Location=Encoder, Tokens=192, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 52.1 | — | |
| ViCropvariant=pure-grad2026.02 | 51.67 | 2.36 | |
| FastVVenue=ECCV’24, Location=LLM, Tokens=128, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 50.6 | — | |
| PyramidDropVenue=CVPR’25, Location=LLM, Tokens=64, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 50.6 | — | |
| TRIMVenue=arXiv’25, Location=Encoder, Tokens=64, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 50 | — | |
| SAM2026.02 | 49.42 | 3.33 | |
| ToMeVenue=ICLR’23, Location=Encoder, Tokens=128, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 49.1 | — | |
| YOLO2026.02 | 48.84 | 0.35 | |
| CLIP2026.02 | 48.55 | 1.07 | |
| Original2026.02 | 47.8 | 0.17 | |
| FastVVenue=ECCV’24, Location=LLM, Tokens=64, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 47.8 | — | |
| SparseVLMVenue=ICML’25, Location=LLM, Tokens=32, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 46.1 | — | |
| ToMeVenue=ICLR’23, Location=Encoder, Tokens=64, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 45.3 | — | |
| FastVVenue=ECCV’24, Location=LLM, Tokens=32, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 42.5 | — | |
| ToMeVenue=ICLR’23, Location=Encoder, Tokens=32, Base Model=LLaVA-1.5-7B, Retraining Status=without retraining2026.04 | 38.3 | — |