Visual Question Answering on ScienceQA image
75.1AccuracyCSR
Evaluation Results
| Method | Links | |
|---|---|---|
| CSRBackbone=LLaVA-1.5-13B2024.05 | 75.1 | |
| Self-rewardingBackbone=LLaVA-1.5-13B2024.05 | 74.3 | |
| SparseVLM#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 74.3 | |
| FastV#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 74.2 | |
| DART#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 74.2 | |
| ToFuLLM=Vicuna-13B, compression_ratio_r=0.9252024.05 | 74 | |
| PITOMELLM=Vicuna-13B, compression_ratio_r=0.9252024.05 | 74 | |
| ToMeLLM=Vicuna-13B, compression_ratio_r=0.9252024.05 | 73.8 | |
| VisionZip#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 73.8 | |
| MQT-LLaVA#Vision Tokens=144, Backbone=LLaVA-1.5-13B2026.03 | 73.6 | |
| PyramidDrop#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 73.3 | |
| Prumerge+#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 73.3 | |
| CDPruner#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 73.2 | |
| M3#Vision Tokens=144, Backbone=LLaVA-1.5-13B2026.03 | 72.9 | |
| LLaVA-1.5-13BLLM=Vicuna-13B2024.05 | 72.8 | |
| LLaVA-1.5-13B#Vision Tokens=576, Backbone=LLaVA-1.5-13B2026.03 | 72.8 | |
| DivPrune#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 72.8 | |
| VisPruner#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 72.8 | |
| TRIM#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 72.4 | |
| FMVR-LLaVA#Vision Tokens=576, Backbone=LLaVA-1.5-13B2026.03 | 72.4 | |
| FMVR-LLaVA#Vision Tokens=144, Backbone=LLaVA-1.5-13B2026.03 | 72 | |
| LLaVA-1.5-13BModel Size=13B2024.05 | 71.6 | |
| FMVR-LLaVA#Vision Tokens=36, Backbone=LLaVA-1.5-13B2026.03 | 71.5 | |
| FMVR-LLaVA#Vision Tokens=9, Backbone=LLaVA-1.5-13B2026.03 | 70.9 | |
| CSRBackbone=LLaVA-1.5-7B2024.05 | 70.7 | |
| DiffRateLLM=Vicuna-13B, compression_ratio_r=0.9252024.05 | 70.6 | |
| LLaVA-1.5-7BLLM=Vicuna-7B2024.05 | 70.4 | |
| DCTLLM=Vicuna-13B, compression_ratio_r=0.9252024.05 | 70.3 | |
| Self-rewardingBackbone=LLaVA-1.5-7B2024.05 | 69.6 | |
| LLaVA-1.5-7BRetained Tokens=576, Retention Ratio=100%, Base Model=LLaVA-1.5-7B2025.08 | 69.5 | |
| MonkeyLLM=Qwen-7B2024.08 | 69.4 | |
| FMVR-LLaVA#Vision Tokens=1, Backbone=LLaVA-1.5-13B2026.03 | 69.4 | |
| PITOMELLM=Vicuna-7B, compression_ratio_r=0.92024.05 | 69 | |
| FASTLLM=Vicuna-7B2024.08 | 68.9 | |
| HiPruneRetained Tokens=192, Retention Ratio=33.3%, Base Model=LLaVA-1.5-7B2025.08 | 68.9 | |
| HiPruneRetained Tokens=64, Retention Ratio=11.1%, Base Model=LLaVA-1.5-7B2025.08 | 68.9 | |
| HiPrune++Retained Tokens=64, Retention Ratio=11.1%, Base Model=LLaVA-1.5-7B2025.08 | 68.9 | |
| POVIDBackbone=LLaVA-1.5-7B2024.05 | 68.8 | |
| HiPrune++Retained Tokens=192, Retention Ratio=33.3%, Base Model=LLaVA-1.5-7B2025.08 | 68.8 | |
| ToMeLLM=Vicuna-7B, compression_ratio_r=0.92024.05 | 68.7 | |
| mPLUG-Owl2LLM=LLaMA-7B2024.08 | 68.7 | |
| HiPruneRetained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 68.7 | |
| ToFuLLM=Vicuna-7B, compression_ratio_r=0.92024.05 | 68.5 | |
| HiPrune++Retained Tokens=128, Retention Ratio=22.2%, Base Model=LLaVA-1.5-7B2025.08 | 68.5 | |
| HiPruneRetained Tokens=128, Retention Ratio=22.2%, Base Model=LLaVA-1.5-7B2025.08 | 68.3 | |
| Qwen-VL-ChatLLM=Qwen-7B2024.08 | 68.2 | |
| Chain of SpotLLM=Vicuna-7B2024.08 | 68.2 | |
| DARTRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 68 | |
| HiPruneRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 68 | |
| EADPRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 68 | |
| CDPrunerRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 67.9 | |
| DivPruneRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 67.8 | |
| CDPrunerRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67.7 | |
| VisionZipRetained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 67.7 | |
| LLaVA-NeXT-7BRetained Tokens=2,880, Token Reduction Ratio=100%2026.07 | 67.6 | |
| EADPRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67.6 | |
| VisionZipRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 67.5 | |
| PruMerge+Retained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67.5 | |
| DivPruneRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67.5 | |
| DARTRetained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 67.5 | |
| CDPrunerRetained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 67.5 | |
| PruMerge+Retained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 67.4 | |
| EADPRetained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 67.4 | |
| SparseVLMRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 67.3 | |
| VisionZipRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67.3 | |
| HiPruneRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67.3 | |
| FastVRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 67.2 | |
| RLHF-VBackbone=LLaVA-1.5-7B2024.05 | 67.1 | |
| TRIMRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 67.1 | |
| DARTRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67.1 | |
| DivPruneRetained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 67.1 | |
| SparseVLMRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 67 | |
| PDropRetained Tokens=640, Token Reduction Ratio=↓ 77.8%2026.07 | 66.9 | |
| LLaVA-1.5-7BModel Size=7B2024.05 | 66.8 | |
| LLaVA-v1.5LLM=Vicuna-7B2024.08 | 66.8 | |
| PruMerge+Retained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 66.7 | |
| PDropRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 66.5 | |
| DiffRateLLM=Vicuna-7B, compression_ratio_r=0.92024.05 | 66.4 | |
| TRIMRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 66.4 | |
| FastVRetained Tokens=320, Token Reduction Ratio=↓ 88.9%2026.07 | 66.3 | |
| VlfeedbackBackbone=LLaVA-1.5-7B2024.05 | 66.2 | |
| DCTLLM=Vicuna-7B, compression_ratio_r=0.92024.05 | 65.8 | |
| Human-PreferBackbone=LLaVA-1.5-7B2024.05 | 65.8 | |
| TRIMRetained Tokens=160, Token Reduction Ratio=↓ 94.4%2026.07 | 65.7 | |
| InstructBLIPLLM=Vicuna-13B2024.05 | 63.1 | |
| InstructBLIPLLM=Vicuna-13B2024.08 | 63.1 | |
| BLIP-2LLM=Vicuna-7B2024.05 | 61 | |
| BLIP-2LLM=Vicuna-13B2024.08 | 61 | |
| InstructBLIPLLM=Vicuna-7B2024.05 | 60.5 |