Visual Question Answering on VQA Aggregated Subsets
100Average RatioVanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| VanillaSource=Qwen’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=0%2026.06 | 100 | |
| PDrop+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.06 | 93.9 | |
| FastV+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.06 | 91.1 | |
| PDrop+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.06 | 89.6 | |
| PDropSource=CVPR’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.06 | 87.9 | |
| FastV+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.06 | 86.2 | |
| PDrop+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.06 | 84.4 | |
| FastVSource=ECCV’24, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.06 | 82.5 | |
| PDropSource=CVPR’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.06 | 78.9 | |
| FastV+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.06 | 76.1 | |
| FastVSource=ECCV’24, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.06 | 72.5 | |
| PDropSource=CVPR’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.06 | 67.5 | |
| FastVSource=ECCV’24, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.06 | 57.1 |