Visual Question Answering on VQAv2 (test)
83.4VQA AccuracyUniGame
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| UniGameParams=7B2025.11 | 83.4 | — | — | — | — | |
| BLIP3-oParams=8B2025.11 | 83.1 | — | — | — | — | |
| SFTParams=7B2025.11 | 79.5 | — | — | — | — | |
| mPLUG-Owl2Type=Und. Only, #Params=7B2024.08 | 79.4 | — | — | — | — | |
| VILA-UType=Und. and Gen., #Params=7B2024.08 | 79.4 | — | — | — | — | |
| VILA-UType=Und. and Gen., # LLM Params=7B, External pretrained diffusion model=false2024.10 | 79.4 | — | — | — | — | |
| FREE-V-F#Train Params=1.5B, Backbone=ViT-g, LLM=FlanT5-XL2025.06 | 79.1 | — | — | — | 1.71 | |
| FREE-V-O#Train Params=1.6B, Backbone=ViT-g, LLM=OPT-2.7B2025.06 | 79 | — | — | — | 1.77 | |
| BLIP-2 V-F#Train Params=1.2B, Backbone=ViT-g, LLM=FlanT5-XL2025.06 | 78.7 | — | — | — | 1 | |
| LLaVA-v1.5Type=Und. Only, #Params=7B2024.08 | 78.5 | — | — | — | — | |
| LLaVA-v1.5Type=Und. Only, # LLM Params=7B, External pretrained diffusion model=false2024.10 | 78.5 | — | — | — | — | |
| BLIP-2 V-O#Train Params=1.2B, Backbone=ViT-g, LLM=OPT-2.7B2025.06 | 78.5 | — | — | — | 1.07 | |
| Qwen-VL-ChatType=Und. Only, #Params=7B2024.08 | 78.2 | — | — | — | — | |
| Qwen-VL-ChatType=Und. Only, # LLM Params=7B, External pretrained diffusion model=false2024.10 | 78.2 | — | — | — | — | |
| Janus-Pro†Params=7B2025.11 | 78.2 | — | — | — | — | |
| Flamingo80B#Train Params=10.6B2025.06 | 78.1 | — | — | — | — | |
| LeeBLIP#Train Params=1.9B2025.06 | 77.8 | — | — | — | 1.65 | |
| TokenFlow-XLParams=14B2025.11 | 77.6 | — | — | — | — | |
| JanusType=Und. and Gen., # LLM Params=1.3B, External pretrained diffusion model=false2024.10 | 77.3 | — | — | — | — | |
| PABEE-BLIP#Train Params=1.9B2025.06 | 77.1 | — | — | — | 1.39 | |
| VK-OOD2023.02 | 76.8 | — | — | — | — | |
| VinVL-BModel Scale=Base2023.02 | 75.9 | — | — | — | — | |
| DeeBLIP#Train Params=1.9B2025.06 | 75.9 | — | — | — | 1.52 | |
| OFA#Train Params=930M2025.06 | 75.6 | — | — | — | — | |
| LLaVA-v1.5-Phi-1.5Type=Und. Only, #Params=1.3B2024.08 | 75.3 | — | — | — | — | |
| LLaVA-v1.5-Phi-1.5Type=Und. Only, # LLM Params=1.3B, External pretrained diffusion model=false2024.10 | 75.3 | — | — | — | — | |
| Emu3Type=Und. and Gen., #Params=8B2024.08 | 75.1 | — | — | — | — | |
| Emu3-ChatType=Und. Only, # LLM Params=8B, External pretrained diffusion model=false2024.10 | 75.1 | — | — | — | — | |
| Emu3Params=8B2025.11 | 75.1 | — | — | — | — | |
| ChameleonType=Und. and Gen., #Params=34B2024.08 | 74.7 | — | — | — | — | |
| Show-o+Type=Und. and Gen., #Params=1.3B, improved with CLIP-ViT=true2024.08 | 74.7 | — | — | — | — | |
| ALBEF(4M)Pre-train images=4M2023.02 | 74.5 | — | — | — | — | |
| PixelBERT2023.02 | 74.5 | — | — | — | — | |
| DreamLLMType=Und. and Gen., #Params=7B2024.08 | 72.9 | — | — | — | — | |
| DreamLLMType=Und. and Gen., # LLM Params=7B, External pretrained diffusion model=false2024.10 | 72.9 | — | — | — | — | |
| FLAVA2023.02 | 72.8 | — | — | — | — | |
| UNITER-BModel Scale=Base2023.02 | 72.7 | — | — | — | — | |
| BLIP#Train Params=385M2025.06 | 72.1 | — | — | — | — | |
| ALBEF#Train Params=314M2025.06 | 71.5 | — | — | — | — | |
| LLaVA-PhiType=Und. Only, # LLM Params=2.7B, External pretrained diffusion model=false2024.10 | 71.4 | — | — | — | — | |
| SEED-XParams=7B2025.11 | 71.2 | — | — | — | — | |
| Zhou et al. (2017) (Ens.)ensemble=true, features=object proposal2018.02 | 70.92 | 86.65 | 51.13 | 61.75 | — | |
| Teney et al. (2017) (Ens.)ensemble=true, features=object proposal2018.02 | 70.34 | 86.6 | 48.64 | 61.15 | — | |
| ViLT-BModel Scale=Base2023.02 | 70.3 | — | — | — | — | |
| Show-oType=Und. and Gen., #Params=1.3B2024.08 | 69.4 | — | — | — | — | |
| Show-oParams=1.3B2025.11 | 69.4 | — | — | — | — | |
| Baseline + counting moduleensemble=false, features=object proposal2018.02 | 68.41 | 83.56 | 51.39 | 59.11 | — | |
| Baselineensemble=false, features=object proposal2018.02 | 67.78 | 83.21 | 46.6 | 59.2 | — | |
| NEXT-GPTType=Und. and Gen., #Params=13B2024.08 | 66.7 | — | — | — | — | |
| NExT-GPTType=Und. and Gen., # LLM Params=13B, External pretrained diffusion model=true2024.10 | 66.7 | — | — | — | — | |
| LaVITType=Und. and Gen., # LLM Params=7B, External pretrained diffusion model=true2024.10 | 66 | — | — | — | — | |
| ChameleonParams=7B2025.11 | 66 | — | — | — | — | |
| Teney et al. (2017)ensemble=false, features=object proposal2018.02 | 65.67 | 82.2 | 43.9 | 56.26 | — | |
| LiquidParams=7B2025.11 | 63.5 | — | — | — | — | |
| Gemini-Nano-1Type=Und. and Gen., #Params=1.8B2024.08 | 62.7 | — | — | — | — | |
| Gemini-Nano-1Type=Und. and Gen., # LLM Params=1.8B, External pretrained diffusion model=false2024.10 | 62.7 | — | — | — | — | |
| BLIP-2 ViT-g FlanT5xL#Total params=4.1B, Speed=1.00x, Unsupervised=true2025.06 | 62.5 | — | — | — | — | |
| FREE ViT-g FlanT5xL#Total params=4.5B, Speed=1.45x, Unsupervised=true2025.06 | 62.1 | — | — | — | — | |
| Img2LLM 175BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false2022.12 | 61.9 | — | — | — | — | |
| Img2LLM 30BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false2022.12 | 60.4 | — | — | — | — | |
| Img2LLM 66BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false2022.12 | 60.3 | — | — | — | — | |
| Show-oType=Und. and Gen., # LLM Params=1.3B, External pretrained diffusion model=false2024.10 | 59.3 | — | — | — | — | |
| Img2LLM 13BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false2022.12 | 57.3 | — | — | — | — | |
| EmuType=Und. and Gen., #Params=13B2024.08 | 57.2 | — | — | — | — | |
| LeeBLIP#Total params=4.7B, Speed=1.29x, Unsupervised=true2025.06 | 57.1 | — | — | — | — | |
| Img2LLM 6.7BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false2022.12 | 57 | — | — | — | — | |
| Flamingo80B#Total params=80B, Speed=0.05x, Unsupervised=true2025.06 | 56.2 | — | — | — | — | |
| LWMType=Und. and Gen., # LLM Params=7B, External pretrained diffusion model=false2024.10 | 55.8 | — | — | — | — | |
| BLIP-2 ViT-g OPT6.7B#Total params=7.8B, Speed=0.52x, Unsupervised=true2025.06 | 53.7 | — | — | — | — | |
| MuE#Total params=4.7B, Speed=1.36x, Unsupervised=true2025.06 | 53.6 | — | — | — | — | |
| FREE ViT-g OPT2.7B#Total params=4.3B, Speed=1.51x, Unsupervised=true2025.06 | 53.4 | — | — | — | — | |
| EmuType=Und. and Gen., # LLM Params=13B, External pretrained diffusion model=true2024.10 | 52 | — | — | — | — | |
| BLIP-2 ViT-g OPT2.7B#Total params=3.8B, Speed=1.07x, Unsupervised=true2025.06 | 52 | — | — | — | — | |
| Flamingo 9BEnd-to-End Training=true, Shot Number=02022.12 | 51.8 | — | — | — | — | |
| Flamingo9B#Total params=9.3B, Speed=0.44x, Unsupervised=true2025.06 | 51.8 | — | — | — | — | |
| PABEE-BLIP#Total params=4.7B, Speed=1.31x, Unsupervised=true2025.06 | 51.3 | — | — | — | — | |
| IDEFICS-9BType=Und. Only, # LLM Params=8B, External pretrained diffusion model=false2024.10 | 50.9 | — | — | — | — | |
| Flamingo3B#Total params=3.2B, Speed=1.28x, Unsupervised=true2025.06 | 49.4 | — | — | — | — | |
| Flamingo 3BEnd-to-End Training=true, Shot Number=02022.12 | 49.2 | — | — | — | — | |
| VLKD ViT-L/14End-to-End Training=true, Shot Number=02022.12 | 44.5 | — | — | — | — | |
| DeeBLIP#Total params=4.7B, Speed=1.39x, Unsupervised=true2025.06 | 42.8 | — | — | — | — | |
| VLKD ViT-B/16End-to-End Training=true, Shot Number=02022.12 | 39.7 | — | — | — | — |