Multimodal Understanding on LLaVA-Bench
91.9Overall ScoreResDec
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ResDecBase Model=Qwen2.5-VL2026.02 | 91.9 | — | — | — | |
| VISTABase Model=Qwen2.5-VL2026.02 | 91.7 | — | — | — | |
| OPERABase Model=Qwen2.5-VL2026.02 | 91.3 | — | — | — | |
| RegularBase Model=Qwen2.5-VL2026.02 | 91 | — | — | — | |
| VCDBase Model=Qwen2.5-VL2026.02 | 89.5 | — | — | — | |
| DoLaBase Model=Qwen2.5-VL2026.02 | 87.3 | — | — | — | |
| ICDBase Model=Qwen2.5-VL2026.02 | 83.7 | — | — | — | |
| Qwen-VL-ChatLLM=Qwen-7B, Resolution=448 × 4482026.04 | 67.7 | — | — | — | |
| VisionZipRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 67.7 | — | — | — | |
| LRCPRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 67.5 | — | — | — | |
| DenseConnectorLLM=Vicuna-7B2026.04 | 67.4 | — | — | — | |
| TARSBase Model=LLaVA-v1.5-7B2025.07 | 67.2 | 65.9 | 57.8 | 75.1 | |
| VanillaRetained Tokens=576, Backbone=LLaVA-v1.5-7B2026.05 | 66.9 | — | — | — | |
| LLaVA-v1.5-7BBudget=Unpruned2026.04 | 66.8 | — | — | — | |
| CLASPBudget=R=1922026.04 | 66.7 | — | — | — | |
| V2DropRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 66.5 | — | — | — | |
| VIFLLM=Vicuna-7B, Resolution=336 × 3362026.04 | 66.4 | — | — | — | |
| SparseVLMBudget=R=1922026.04 | 66.1 | — | — | — | |
| SparseVLMRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 66.1 | — | — | — | |
| ApETRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 66.1 | — | — | — | |
| PDropRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 65.8 | — | — | — | |
| CLASPBudget=R=1282026.04 | 65.6 | — | — | — | |
| MMFuserLLM=Vicuna-7B2026.04 | 65.5 | — | — | — | |
| VisionZipRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 64.8 | — | — | — | |
| LRCPRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 64.8 | — | — | — | |
| DPOBase Model=LLaVA-v1.5-7B2025.07 | 64.6 | 63.5 | 55.4 | 72.8 | |
| ResDecBase Model=InstructBLIP2026.02 | 64.4 | — | — | — | |
| ApETRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 64.2 | — | — | — | |
| CLASPBudget=R=642026.04 | 64 | — | — | — | |
| VISTABase Model=InstructBLIP2026.02 | 63.5 | — | — | — | |
| OPERABase Model=InstructBLIP2026.02 | 63.1 | — | — | — | |
| LRCPRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 63.1 | — | — | — | |
| ApETRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 63 | — | — | — | |
| LLaVA-v1.5-7B2025.07 | 62.9 | 61.8 | 53.6 | 71.2 | |
| V2DropRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 62.9 | — | — | — | |
| VisionZipRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 62.9 | — | — | — | |
| SparseVLMBudget=R=1282026.04 | 62.7 | — | — | — | |
| SparseVLMRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 62.7 | — | — | — | |
| ResDecBase Model=LLaVA-1.52026.02 | 62.6 | — | — | — | |
| V2DropRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 62.4 | — | — | — | |
| MemVRBase Model=LLaVA-1.52026.02 | 62.2 | — | — | — | |
| PDropRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 61.9 | — | — | — | |
| RegularBase Model=LLaVA-1.52026.02 | 61.8 | — | — | — | |
| LLaVA-v1.5LLM=Vicuna-7B, Resolution=336 × 3362026.04 | 61.8 | — | — | — | |
| ONLYBase Model=LLaVA-1.52026.02 | 61.6 | — | — | — | |
| OPERABase Model=LLaVA-1.52026.02 | 61.4 | — | — | — | |
| VCDBase Model=InstructBLIP2026.02 | 61.1 | — | — | — | |
| VISTABase Model=LLaVA-1.52026.02 | 60.9 | — | — | — | |
| DoLaBase Model=LLaVA-1.52026.02 | 60.7 | — | — | — | |
| mPLUG-Owl2LLM=LLaMA 2-7B, Resolution=448 × 4482026.04 | 59.9 | — | — | — | |
| RegularBase Model=InstructBLIP2026.02 | 59.8 | — | — | — | |
| InstructBLIPLLM=Vicuna-7B, Resolution=224 × 2242026.04 | 59.8 | — | — | — | |
| PDropRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 59.2 | — | — | — | |
| VCDBase Model=LLaVA-1.52026.02 | 58.5 | — | — | — | |
| SparseVLMBudget=R=642026.04 | 57.5 | — | — | — | |
| SparseVLMRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 57.5 | — | — | — | |
| DoLaBase Model=InstructBLIP2026.02 | 56.6 | — | — | — | |
| ICDBase Model=LLaVA-1.52026.02 | 55.1 | — | — | — | |
| FastVRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 52 | — | — | — | |
| ICDBase Model=InstructBLIP2026.02 | 51 | — | — | — | |
| FastVRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 49.4 | — | — | — | |
| Fed-CMPSetting=MCross2026.03 | 48.2 | — | — | — | |
| Mantis-8B-FuyuLLM=Fuyu-8B, Resolution=1024 × 10242026.04 | 46.8 | — | — | — | |
| Fed-CMPsetting=MI-I2026.03 | 46.6 | — | — | — | |
| FedAvg-DARESetting=MCross2026.03 | 46.3 | — | — | — | |
| FastVRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 46.1 | — | — | — | |
| MiniGPT-v2LLM=LLaMA 2-7B, Resolution=448 × 4482026.04 | 45.1 | — | — | — | |
| IDEFICS-9BLLM=LLaMA 2-7B2026.04 | 45 | — | — | — | |
| Fed-CMPSetting=M_Joint2026.03 | 44.8 | — | — | — | |
| FedAvg-TIEsSetting=MCross2026.03 | 43.7 | — | — | — | |
| FedAvg-DAREsetting=MI-I2026.03 | 42.5 | — | — | — | |
| MOONSetting=MCross2026.03 | 42.1 | — | — | — | |
| FedAvg-TASetting=MCross2026.03 | 41.9 | — | — | — | |
| FedAvg-DARESetting=M_Joint2026.03 | 41.6 | — | — | — | |
| FedAvg-TIEsSetting=M_Joint2026.03 | 41.5 | — | — | — | |
| FedAvg-TIEssetting=MI-I2026.03 | 41.3 | — | — | — | |
| FedAvgSetting=MCross2026.03 | 41.3 | — | — | — | |
| MOONsetting=MI-I2026.03 | 41.2 | — | — | — | |
| FedAvg-TAsetting=MI-I2026.03 | 40.7 | — | — | — | |
| MOONSetting=M_Joint2026.03 | 40.6 | — | — | — | |
| FedAvg-TASetting=M_Joint2026.03 | 40.2 | — | — | — | |
| FedAvgsetting=MI-I2026.03 | 40 | — | — | — | |
| FedAvgSetting=M_Joint2026.03 | 39.8 | — | — | — | |
| FedProxSetting=MCross2026.03 | 39.8 | — | — | — | |
| FedAdamsetting=MI-I2026.03 | 39.2 | — | — | — | |
| FedAdamSetting=M_Joint2026.03 | 39.1 | — | — | — | |
| FedAdamSetting=MCross2026.03 | 38.5 | — | — | — | |
| FedProxsetting=MI-I2026.03 | 38.4 | — | — | — | |
| FedProxSetting=M_Joint2026.03 | 37.5 | — | — | — | |
| Local TrainingSetting=M_Joint2026.03 | 37.3 | — | — | — | |
| Local TrainingSetting=MCross2026.03 | 36.7 | — | — | — | |
| Local Trainingsetting=MI-I2026.03 | 35.9 | — | — | — | |
| LLaMA-Adapter-v2LLM=LLaMA 2-7B2026.04 | 32.7 | — | — | — | |
| Qwen-VLLLM=Qwen-7B, Resolution=448 × 4482026.04 | 12.9 | — | — | — |