Visual Question Answering on MSTAR
40.9AccuracyM³
Evaluation Results
| Method | Links | |
|---|---|---|
| M³FLOPs Savings=8.0×2026.03 | 40.9 | |
| LLaVA-OVFLOPs Savings=1.0×2026.03 | 40.6 | |
| PyramidDropFLOPs Savings=4.2×2026.03 | 40.1 | |
| VISORFLOPs Savings=8.6×2026.03 | 40.1 | |
| DownsampleFLOPs Savings=4.0×2026.03 | 40 | |
| VisionZip†FLOPs Savings=5.7×2026.03 | 39.3 | |
| HiREDFLOPs Savings=5.0×2026.03 | 39.2 | |
| VisionZipFLOPs Savings=5.7×2026.03 | 38.9 | |
| SparseVLMFLOPs Savings=4.5×2026.03 | 38.5 | |
| VISOR-TRFLOPs Savings=18×2026.03 | 38.5 | |
| VisPrunerFLOPs Savings=5.7×2026.03 | 38.1 |