List your company

Visual Question Answering on VQA lite v2

80AccuracyVanilla
36.00847.42958.8570.271Jun 10, 2025Aug 9, 2025Oct 9, 2025Dec 9, 2025Feb 8, 2026Apr 10, 2026Jun 10, 2026

Evaluation results

MethodLinks
VanillaAverage Token Reduction=0.0%, Backbone=Qwen3-VL-8B2026.0680
VanillaSource=Qwen’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=0%2026.0679.6
VanillaSource=Qwen’25, Token Reduction=0%, Backbone=Qwen2.5-VL-7B2026.0678.6
PDrop+RerouteSource=Ours, Token Reduction=66.7%, Backbone=Qwen2.5-VL-7B2026.0677.6
SECONDModel=LLaVA-Next (CLIP-336), LLM=Vicuna-7B, CD=✓2025.0677.5
FastV+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.0677.2
SECONDModel=LLaVA-Next (CLIP-336), LLM=Vicuna-7B, CD=✗2025.0676.5
baselineModel=LLaVA-Next (CLIP-336), LLM=Vicuna-7B, CD=✗2025.0676.4
FastV+RerouteSource=Ours, Token Reduction=66.7%, Backbone=Qwen2.5-VL-7B2026.0676.3
PDropSource=CVPR’25, Token Reduction=66.7%, Backbone=Qwen2.5-VL-7B2026.0675.8
PDrop+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.0675.4
SECONDModel=LLaVA-OneVision (SigLIP-384), LLM=Qwen2-0.5B, CD=✓2025.0675.1
PDrop+RerouteAverage Token Reduction=66.7%, Backbone=Qwen3-VL-8B2026.0675.1
baselineModel=LLaVA-OneVision (SigLIP-384), LLM=Qwen2-0.5B, CD=✗2025.0674.6
SECONDModel=LLaVA-Next (CLIP-336), LLM=Mistral-7B, CD=✓2025.0674.5
PDrop+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.0674.4
PDropAverage Token Reduction=66.7%, Backbone=Qwen3-VL-8B2026.0674.2
PDrop+RerouteSource=Ours, Token Reduction=77.8%, Backbone=Qwen2.5-VL-7B2026.0674.1
SECONDModel=LLaVA-Next (CLIP-336), LLM=Mistral-7B, CD=✗2025.0673.6
SECONDModel=LLaVA-OneVision (SigLIP-384), LLM=Qwen2-0.5B, CD=✗2025.0673.6
VCDModel=LLaVA-Next (CLIP-336), LLM=Vicuna-7B, CD=✓2025.0672.9
FastVSource=ECCV’24, Token Reduction=66.7%, Backbone=Qwen2.5-VL-7B2026.0672.9
FastV+RerouteSource=Ours, Token Reduction=77.8%, Backbone=Qwen2.5-VL-7B2026.0672.4
FastV+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.0672.3
baselineModel=LLaVA-Next (CLIP-336), LLM=Mistral-7B, CD=✗2025.0672
FastV+RerouteAverage Token Reduction=66.7%, Backbone=Qwen3-VL-8B2026.0671.8
PDrop+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.0670.8
FastVAverage Token Reduction=66.7%, Backbone=Qwen3-VL-8B2026.0670.6
VCDModel=LLaVA-Next (CLIP-336), LLM=Mistral-7B, CD=✓2025.0670.1
PDrop+RerouteAverage Token Reduction=77.8%, Backbone=Qwen3-VL-8B2026.0669.3
PDrop+RerouteSource=Ours, Token Reduction=88.9%, Backbone=Qwen2.5-VL-7B2026.0668.6
PDropSource=CVPR’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.0668.6
PDropSource=CVPR’25, Token Reduction=77.8%, Backbone=Qwen2.5-VL-7B2026.0667.6
SECONDModel=Yi-VL (CLIP-448), LLM=Yi-6B, CD=✓2025.0665.3
FastV+RerouteSource=Ours, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.0664.5
baselineModel=Yi-VL (CLIP-448), LLM=Yi-6B, CD=✗2025.0664.3
FastVSource=ECCV’24, Token Reduction=77.8%, Backbone=Qwen2.5-VL-7B2026.0664.2
SECONDModel=Yi-VL (CLIP-448), LLM=Yi-6B, CD=✗2025.0663.6
FastVAverage Token Reduction=77.8%, Backbone=Qwen3-VL-8B2026.0663.4
FastVSource=ECCV’24, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=66.7%2026.0662.8
PDropAverage Token Reduction=77.8%, Backbone=Qwen3-VL-8B2026.0662.5
FastV+RerouteAverage Token Reduction=77.8%, Backbone=Qwen3-VL-8B2026.0662.1
VCDModel=Yi-VL (CLIP-448), LLM=Yi-6B, CD=✓2025.0661.9
PDrop+RerouteAverage Token Reduction=88.9%, Backbone=Qwen3-VL-8B2026.0660.6
PDropSource=CVPR’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.0658.6
PDropSource=CVPR’25, Token Reduction=88.9%, Backbone=Qwen2.5-VL-7B2026.0658.2
VCDModel=LLaVA-OneVision (SigLIP-384), LLM=Qwen2-0.5B, CD=✓2025.0655
FastVSource=ECCV’24, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=77.8%2026.0653.1
FastV+RerouteSource=Ours, Token Reduction=88.9%, Backbone=Qwen2.5-VL-7B2026.0652.3
PDropAverage Token Reduction=88.9%, Backbone=Qwen3-VL-8B2026.0651.6
PDropSource=CVPR’25, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.0647.3
FastV+RerouteAverage Token Reduction=88.9%, Backbone=Qwen3-VL-8B2026.0642.7
FastVAverage Token Reduction=88.9%, Backbone=Qwen3-VL-8B2026.0640.8
FastVSource=ECCV’24, Backbone=Qwen3.5-9B-Hybrid, Average Gated Attention Token Reduction=88.9%2026.0638.9
FastVSource=ECCV’24, Token Reduction=88.9%, Backbone=Qwen2.5-VL-7B2026.0637.7