Multimodal Evaluation on MME (Perception (P) and Sum)
1,510Perception Score (P)LLaVA-1.5
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLaVA-1.5LLM=Vicuna-7B, Retain Ratio (%)=100%2026.05 | 1,510 | 1,862 | |
| VEN-VLLLM=Qwen3-1.7B, Retain Ratio (%)=∼ 7.5%2026.05 | 1,504 | 1,984 | |
| Qwen-VL-ChatLLM=Qwen-7B, Retain Ratio (%)=100%2026.05 | 1,488 | 1,848 | |
| LLaVA-MiniLLM=Vicuna-7B, Retain Ratio (%)=∼ 10%2026.05 | 1,477 | — | |
| PruMerge+LLM=Vicuna-7B, Retain Ratio (%)=∼ 25%2026.05 | 1,462 | — | |
| TinyLLaVA†LLM=Qwen2.5-1.5B, Retain Ratio (%)=100%2026.05 | 1,450 | — | |
| VEN-VLLLM=Qwen2.5-1.5B, Retain Ratio (%)=∼ 10%2026.05 | 1,401 | 1,736 | |
| LLaVA-MODLLM=Qwen2-1.5B, Retain Ratio (%)=100%2026.05 | 1,384 | — | |
| LLaVA-MODLLM=Qwen1.5-1.8B, Retain Ratio (%)=100%2026.05 | 1,352 | — | |
| PruMergeLLM=Vicuna-7B, Retain Ratio (%)=∼ 5%2026.05 | 1,350 | — | |
| Mini-Gemini-2BLLM=Gemma-2B, Retain Ratio (%)=100%2026.05 | 1,341 | 1,653 | |
| VEN-VLLLM=Qwen3-0.6B, Retain Ratio (%)=∼ 7.5%2026.05 | 1,320 | 1,614 | |
| Imp-2BLLM=Qwen1.5-1.8B, Retain Ratio (%)=100%2026.05 | 1,304 | — | |
| Bunny-2BLLM=Qwen1.5-1.8B, Retain Ratio (%)=100%2026.05 | 1,300 | — | |
| MoE-LLaVA-2BLLM=Qwen-1.5-1.8B, Retain Ratio (%)=100%2026.05 | 1,292 | — | |
| TinyLLaVA†LLM=Qwen2.5-0.5B, Retain Ratio (%)=100%2026.05 | 1,230 | — | |
| TinyLLaVA†LLM=Qwen1.5-1.8B, Retain Ratio (%)=100%2026.05 | 1,224 | — | |
| TinyLLaVA†LLM=Qwen1.5-0.5B, Retain Ratio (%)=100%2026.05 | 1,196 | — | |
| FastVLLM=Vicuna-7B, Retain Ratio (%)=∼ 10%2026.05 | — | 1,256 | |
| InternVL2-2BLLM=InternLM2-1.8B, Retain Ratio (%)=100%2026.05 | — | 1,863 | |
| MiniCPM-V2LLM=MiniCPM-2.4B, Retain Ratio (%)=100%2026.05 | — | 1,788 | |
| PyramidDropLLM=Vicuna-7B, Retain Ratio (%)=∼ 10%2026.05 | — | 1,561 | |
| Qwen2-VL-2BLLM=Qwen2-1.5B, Retain Ratio (%)=100%2026.05 | — | 1,890 | |
| SparseVLMLLM=Vicuna-7B, Retain Ratio (%)=∼ 10%2026.05 | — | 1,589 | |
| VisionZipLLM=Vicuna-7B, Retain Ratio (%)=∼ 10%2026.05 | — | 1,676 |