Visual Question Answering on VQA v2 (test)
86.1AccuracyLyrics
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| LyricsShot=4-shot2023.12 | 86.1 | — | — | — | — | — | — | — | |
| LyricsShot=1-shot2023.12 | 85.2 | — | — | — | — | — | — | — | |
| PaLI-17BType=Specialists, Evaluation Protocol=Fine-tuned2024.10 | 84.3 | — | — | — | — | — | — | — | |
| LLaVA-NeXT-13BBackbone=LLaVA-NeXT-13B, Retained Tokens=2880, Pruning Ratio=0%2026.02 | 82.3 | — | — | — | — | — | — | — | |
| LLaVA-1.5+CoSLanguage=Vicuna-13B2024.03 | 81.8 | — | — | — | — | — | — | — | |
| GIT2Type=Specialists, Evaluation Protocol=Fine-tuned2024.10 | 81.7 | — | — | — | — | — | — | — | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=960, Pruning Ratio=66.7%2026.02 | 81.6 | — | — | — | — | — | — | — | |
| Qwen-VLShot=4-shot2023.12 | 81.3 | — | — | — | — | — | — | — | |
| FSRBackbone=LLaVA-NeXT-13B, Retained Tokens=960, Pruning Ratio=66.7%2026.02 | 81.3 | — | — | — | — | — | — | — | |
| ShikraShot=4-shot2023.12 | 81.2 | — | — | — | — | — | — | — | |
| LyricsShot=0-shot2023.12 | 81.2 | — | — | — | — | — | — | — | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 81 | — | — | — | — | — | — | — | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=960, Pruning Ratio=66.7%2026.02 | 80.8 | — | — | — | — | — | — | — | |
| LLaVA-1.5+CoSLanguage=Vicuna-7B2024.03 | 80.7 | — | — | — | — | — | — | — | |
| ShareGPT4V-7BLanguage Model=Vicuna-7B2023.11 | 80.6 | — | — | — | — | — | — | — | |
| FSRBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 80.6 | — | — | — | — | — | — | — | |
| Qwen-VLShot=1-shot2023.12 | 80.5 | — | — | — | — | — | — | — | |
| DivPruneBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 80.4 | — | — | — | — | — | — | — | |
| MonkeyLanguage=Qwen-7B2024.03 | 80.3 | — | — | — | — | — | — | — | |
| LLaVA-1.5Language Model=Vicuna-13B2023.11 | 80 | — | — | — | — | — | — | — | |
| LLaVA-1.5Language=Vicuna-13B2024.03 | 80 | — | — | — | — | — | — | — | |
| LLaVA-13BNumber of Parameters=13B2024.05 | 80 | — | — | — | — | — | — | — | |
| LLaVA-1.5-13BBackbone=LLaVA-1.5-13B, Retained Tokens=576, Pruning Ratio=0%2026.02 | 80 | — | — | — | — | — | — | — | |
| JanusFlowLLM Param=1.3B, Type=Unified2024.11 | 79.8 | — | — | — | — | — | — | — | |
| ShikraShot=1-shot2023.12 | 79.7 | — | — | — | — | — | — | — | |
| VisionZipBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 79.7 | — | — | — | — | — | — | — | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 79.7 | — | — | — | — | — | — | — | |
| Qwen-VLLanguage=Qwen-7B2024.03 | 79.5 | — | — | — | — | — | — | — | |
| ArcanaType=Generalists, LLM=Vicuna (7B), Evaluation Protocol=In-domain, MM-LoRA trained during pretrain=true2024.10 | 79.5 | — | — | — | — | — | — | — | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 79.5 | — | — | — | — | — | — | — | |
| mPLUG-Owl2Language=LLaMA-7B2024.03 | 79.4 | — | — | — | — | — | — | — | |
| mPLUG-Owl2Type=Generalists, LLM=LLaMA (7B), Evaluation Protocol=In-domain2024.10 | 79.4 | — | — | — | — | — | — | — | |
| VILA-ULLM Param=7B, Type=Unified2024.11 | 79.4 | — | — | — | — | — | — | — | |
| FastVBackbone=LLaVA-NeXT-13B, Retained Tokens=640, Pruning Ratio=77.8%2026.02 | 79.4 | — | — | — | — | — | — | — | |
| ArcanaType=Generalists, LLM=Vicuna (7B), Evaluation Protocol=In-domain, MM-LoRA trained during pretrain=false2024.10 | 79.2 | — | — | — | — | — | — | — | |
| LLaVALanguage Model=Vicuna-7B2023.11 | 79 | — | — | — | — | — | — | — | |
| IVM-Enhanced LLaVA-7BNumber of Parameters=14B2024.05 | 79 | — | — | — | — | — | — | — | |
| ChatSearcher2024.10 | 78.9 | — | — | — | — | — | — | — | |
| Qwen-VLLanguage Model=Qwen-7B2023.11 | 78.8 | — | — | — | — | — | — | — | |
| Qwen-7BNumber of Parameters=7B2024.05 | 78.8 | — | — | — | — | — | — | — | |
| FSRBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 78.8 | — | — | — | — | — | — | — | |
| Qwen-VL2024.10 | 78.8 | — | — | — | — | — | — | — | |
| LocVLM-LLLM=7B, Visual Scale=336, Zero-Shot=false2024.04 | 78.6 | — | — | — | — | — | — | — | |
| FSRBackbone=LLaVA-1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.02 | 78.6 | — | — | — | — | — | — | — | |
| LLaVA-1.5Language Model=Vicuna-7B2023.11 | 78.5 | — | — | — | — | — | — | — | |
| LLaVA-1.5Language=Vicuna-7B2024.03 | 78.5 | — | — | — | — | — | — | — | |
| LLaVA-v1.5Type=Generalists, LLM=Vicuna (7B), Evaluation Protocol=In-domain2024.10 | 78.5 | — | — | — | — | — | — | — | |
| LLaVA-7BNumber of Parameters=7B2024.05 | 78.5 | — | — | — | — | — | — | — | |
| LLaVA-v1.5LLM Param=7B, Type=Und. Only2024.11 | 78.5 | — | — | — | — | — | — | — | |
| VanillaBase Model=LLaVA1.5-7B, Pruning Family=Baseline2026.02 | 78.5 | — | — | 100 | — | — | — | — | |
| LLaVA-1.5-7BVersion=1.5, Parameters=7B2024.10 | 78.5 | — | — | — | — | — | — | — | |
| LLaVA-v1.5LLM=7B, Visual Scale=336, Zero-Shot=false2024.04 | 78.4 | — | — | — | — | — | — | — | |
| CDPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.02 | 78.4 | — | — | — | — | — | — | — | |
| Qwen-VL-ChatLanguage Model=Qwen-7B2023.11 | 78.2 | — | — | — | — | — | — | — | |
| Qwen-VL-ChatLanguage=Qwen-7B2024.03 | 78.2 | — | — | — | — | — | — | — | |
| Qwen-VLShot=0-shot2023.12 | 78.2 | — | — | — | — | — | — | — | |
| Qwen-VL-ChatType=Generalists, LLM=Qwen (7B), Evaluation Protocol=In-domain2024.10 | 78.2 | — | — | — | — | — | — | — | |
| LISA-Enhanced LLaVA-7BNumber of Parameters=20B2024.05 | 78.2 | — | — | — | — | — | — | — | |
| Qwen-VL-ChatLLM Param=7B, Type=Und. Only2024.11 | 78.2 | — | — | — | — | — | — | — | |
| Qwen-VL-Chat2024.10 | 78.2 | — | — | — | — | — | — | — | |
| VisPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.02 | 78.1 | — | — | — | — | — | — | — | |
| DivPruneBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 78.1 | — | — | — | — | — | — | — | |
| FSRBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 78 | — | — | — | — | — | — | — | |
| CDPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 77.7 | — | — | — | — | — | — | — | |
| METERFine-tuning protocol=original fine-tuned2023.05 | 77.5 | — | — | — | — | — | — | — | |
| ShikraLanguage Model=Vicuna-13B2023.11 | 77.4 | — | — | — | — | — | — | — | |
| ShikraLanguage=Vicuna-13B2024.03 | 77.4 | — | — | — | — | — | — | — | |
| ShikraLLM=7B, Visual Scale=224, Zero-Shot=false2024.04 | 77.4 | — | — | — | — | — | — | — | |
| ShikraShot=0-shot2023.12 | 77.4 | — | — | — | — | — | — | — | |
| ShikraType=Generalists, LLM=Vicuna (7B), Evaluation Protocol=In-domain2024.10 | 77.4 | — | — | — | — | — | — | — | |
| JanusLLM Param=1.3B, Type=Unified2024.11 | 77.3 | — | — | — | — | — | — | — | |
| DivPruneBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 77.1 | — | — | — | — | — | — | — | |
| VisPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 76.9 | — | — | — | — | — | — | — | |
| METER + PuMerFine-tuning protocol=PuMer fine-tuned2023.05 | 76.8 | 1.82 | 38 | — | — | — | — | — | |
| VisionZipBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 76.8 | — | — | — | — | — | — | — | |
| FSRBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 76.8 | — | — | — | — | — | — | — | |
| VisionZipBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 76.8 | — | — | — | — | — | — | — | |
| CDPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 76.7 | — | — | — | — | — | — | — | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 76.7 | — | — | — | — | — | — | — | |
| MMaDAParams=8B, Base model=LLaDA, Architecture=Discrete Diff., Data scale=Unknown, Resolution=512 × 5122025.05 | 76.7 | — | — | — | — | — | — | — | |
| LLaVA-v1.5-Phi-1.5LLM Param=1.3B, Type=Und. Only2024.11 | 75.3 | — | — | — | — | — | — | — | |
| FastVBackbone=LLaVA-1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.02 | 75.3 | — | — | — | — | — | — | — | |
| DivPruneBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 75.2 | — | — | — | — | — | — | — | |
| Emu3-ChatLLM Param=8B, Type=Und. Only2024.11 | 75.1 | — | — | — | — | — | — | — | |
| DREAMLLM-7B*Shots (k)=4, Capability=Comprehension & Creation, Training Data=LLaVA-1.5 SFT2023.09 | 74.4 | — | — | — | — | — | — | — | |
| VisPrunerBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 73.9 | — | — | — | — | — | — | — | |
| DREAMLLM-7B*Shots (k)=2, Capability=Comprehension & Creation, Training Data=LLaVA-1.5 SFT2023.09 | 73.8 | — | — | — | — | — | — | — | |
| DREAMLLM-7B*Shots (k)=8, Capability=Comprehension & Creation, Training Data=LLaVA-1.5 SFT2023.09 | 73.8 | — | — | — | — | — | — | — | |
| VisionZipBackbone=LLaVA-1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.02 | 73.7 | — | — | — | — | — | — | — | |
| DreamLLM+LLM Param=7B, Type=Unified2024.11 | 72.9 | — | — | — | — | — | — | — | |
| BLIP2Shot=4-shot2023.12 | 72.5 | — | — | — | — | — | — | — | |
| VisionZipBase Model=LLaVA1.5-7B, Pruning Family=Vision Encoder-Side Pruning2026.02 | 72.4 | — | — | 92.2 | — | — | — | — | |
| PruMerge+Base Model=LLaVA1.5-7B, Pruning Family=Vision Encoder-Side Pruning2026.02 | 71.3 | — | — | 90.8 | — | — | — | — | |
| MudditParams=1B, Base model=Meissonic, Architecture=Discrete Diff, Data scale=16M, Resolution=1024 × 10242025.05 | 70.2 | — | — | — | — | — | — | — | |
| FastVBackbone=LLaVA-NeXT-13B, Retained Tokens=320, Pruning Ratio=88.9%2026.02 | 69.8 | — | — | — | — | — | — | — | |
| ViLTFine-tuning protocol=original fine-tuned2023.05 | 69.5 | — | — | — | — | — | — | — | |
| Show-OParams=1.3B, Base model=Phi-1.5, Architecture=AR + Discrete Diff, Data scale=35M, Resolution=512 × 5122025.05 | 69.4 | — | — | — | — | — | — | — | |
| PoolingBase Model=LLaVA1.5-7B, Pruning Family=LLM Single-Layer Pruning2026.02 | 69.1 | — | — | 88 | — | — | — | — | |
| ViLT + PuMerFine-tuning protocol=PuMer fine-tuned2023.05 | 68.9 | 1.76 | 45 | — | — | — | — | — | |
| SparseVLMBase Model=LLaVA1.5-7B, Pruning Family=LLM Multi-Layer Pruning2026.02 | 68.2 | — | — | 86.9 | — | — | — | — |