Visual Question Answering on VQA 2.0 (val)
86.1Accuracy (Overall)Molmo2-8B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Molmo2-8BParameter Scale=8B2026.05 | 86.1 | — | — | — | — | |
| Molmo2-4BModel=Molmo2-4B2026.05 | 85.3 | — | — | — | — | |
| Molmo2-4BParameter Scale=4B2026.05 | 85.3 | — | — | — | — | |
| PerceptionLM-8BParameter Scale=8B2026.05 | 84 | — | — | — | — | |
| DeepSeek-VL2-16B-A2.4BModel=DeepSeek-VL2-16B-A2.4B2026.05 | 83.7 | — | — | — | — | |
| MolmoE-8B-A1BModel=MolmoE-8B-A1B2026.05 | 82.8 | — | — | — | — | |
| Zamba2-VL-7BLanguage-backbone scale=7–8B2026.05 | 82.8 | — | — | — | — | |
| Zamba2-VL-7BParameter Scale=7B2026.05 | 82.8 | — | — | — | — | |
| Qwen3-VL-8BParameter Scale=8B2026.05 | 82.5 | — | — | — | — | |
| Qwen3-VL-4BModel=Qwen3-VL-4B2026.05 | 80.7 | — | — | — | — | |
| Qwen3-VL-4BParameter Scale=4B2026.05 | 80.7 | — | — | — | — | |
| Qwen3.5-4BModel=Qwen3.5-4B2026.05 | 80.4 | — | — | — | — | |
| ZAYA1-VL-8B-A1BModel=ZAYA1-VL-8B-A1B2026.05 | 80 | — | — | — | — | |
| PerceptionLM-1BParameter Scale=1B2026.05 | 80 | — | — | — | — | |
| Qwen2.5-VL-3BModel=Qwen2.5-VL-3B2026.05 | 79.6 | — | — | — | — | |
| Zamba2-VL-2.7BLanguage-backbone scale=2–4B2026.05 | 79.6 | — | — | — | — | |
| Zamba2-VL-2.7BParameter Scale=2.7B2026.05 | 79.6 | — | — | — | — | |
| Cobra-8BLanguage-backbone scale=7–8B2026.05 | 79.2 | — | — | — | — | |
| Qwen3-VL-2BModel=Qwen3-VL-2B2026.05 | 78.8 | — | — | — | — | |
| Qwen3-VL-2BParameter Scale=2B2026.05 | 78.8 | — | — | — | — | |
| InternVL3.5-8BParameter Scale=8B2026.05 | 78.6 | — | — | — | — | |
| InternVL3.5-20B-A4BModel=InternVL3.5-20B-A4B2026.05 | 78.4 | — | — | — | — | |
| Qwen3.5-2BModel=Qwen3.5-2B2026.05 | 78.3 | — | — | — | — | |
| Zamba2-VL-1.2BLanguage-backbone scale=~1B2026.05 | 78 | — | — | — | — | |
| Zamba2-VL-1.2BParameter Scale=1.2B2026.05 | 78 | — | — | — | — | |
| Cobra-3.5BLanguage-backbone scale=2–4B2026.05 | 77.8 | — | — | — | — | |
| PLM-3BModel=PLM-3B2026.05 | 77.3 | — | — | — | — | |
| PerceptionLM-3BParameter Scale=3B2026.05 | 76.9 | — | — | — | — | |
| VL-Mamba2.8BLanguage-backbone scale=2–4B2026.05 | 76.6 | — | — | — | — | |
| SimVLMhugecategory=Discriminative2021.08 | 76.5 | — | — | — | — | |
| InternVL3.5-4BModel=InternVL3.5-4B2026.05 | 76.4 | — | — | — | — | |
| InternVL3.5-4BParameter Scale=4B2026.05 | 76.4 | — | — | — | — | |
| Mamba-VL-2.8BLanguage-backbone scale=2–4B2026.05 | 76.1 | — | — | — | — | |
| SimVLMlargecategory=Discriminative2021.08 | 76 | — | — | — | — | |
| SimVLMhugecategory=Generative2021.08 | 75.5 | — | — | — | — | |
| SimVLMlargecategory=Generative2021.08 | 75.2 | — | — | — | — | |
| Pythia-VL-2.8BLanguage-backbone scale=2–4B2026.05 | 75.1 | — | — | — | — | |
| Mamba-VL-1.4BLanguage-backbone scale=~1B2026.05 | 74.5 | — | — | — | — | |
| SimVLMbasecategory=Discriminative2021.08 | 73.8 | — | — | — | — | |
| OscarImage Repr.=Feature Emb., Few-shot=false2021.09 | 73.8 | — | — | — | — | |
| InternVL3.5-2BModel=InternVL3.5-2B2026.05 | 73.6 | — | — | — | — | |
| Pythia-VL-1.4BLanguage-backbone scale=~1B2026.05 | 73.6 | — | — | — | — | |
| InternVL3.5-2BParameter Scale=2B2026.05 | 73.6 | — | — | — | — | |
| SimVLMbasecategory=Generative2021.08 | 73.2 | — | — | — | — | |
| Pythia-VL-1BLanguage-backbone scale=~1B2026.05 | 72.3 | — | — | — | — | |
| Mamba-VL-0.8BLanguage-backbone scale=~1B2026.05 | 71.7 | — | — | — | — | |
| CFR2021.10 | 69.7 | — | — | — | — | |
| InternVL3.5-1BParameter Scale=1B2026.05 | 69.6 | — | — | — | — | |
| MCANed-6#Params (x10^6)=56, FLOPs (x10^9)=2.82019.06 | 67.23 | — | — | — | 0.01 | |
| MCAN2021.10 | 67.2 | — | — | — | — | |
| REGAT2021.10 | 67.2 | — | — | — | — | |
| Pythia2021.10 | 66.3 | — | — | — | — | |
| DFAF2021.10 | 66.2 | — | — | — | — | |
| BAN-8#Params (x10^6)=79, FLOPs (x10^9)=3.32019.06 | 66.04 | — | — | — | 0.08 | |
| BAN2-CTItraining_type=student model2019.09 | 66 | — | — | — | — | |
| BAN2021.10 | 66 | — | — | — | — | |
| CTI2021.10 | 66 | — | — | — | — | |
| MFH#Params (x10^6)=116, FLOPs (x10^9)=4.42019.06 | 65.65 | — | — | — | 0.05 | |
| BAN22019.09 | 65.6 | — | — | — | — | |
| HAN2021.10 | 65.5 | — | — | — | — | |
| MuRel2021.10 | 65.1 | — | — | — | — | |
| XNMsexpert layout=no2018.12 | 64.7 | — | — | — | — | |
| UpDn2019.05 | 63.5 | 81.2 | 42.1 | 55.7 | — | |
| Up-Downexpert layout=no, reported_in_original_paper=true2018.12 | 63.2 | — | — | — | — | |
| Bottom-up2019.09 | 63.2 | — | — | — | — | |
| Baseline architecture2019.06 | 63.1 | — | — | — | — | |
| UpDn+AdvReg.2019.05 | 62.8 | 79.8 | 42.4 | 55.2 | — | |
| UpDn+HINTExpl.=HAT2019.05 | 62.5 | 80.5 | 41.8 | 54 | — | |
| UpDn+SCRExpl.=QA2019.05 | 62.3 | 77.4 | 40.9 | 56.5 | — | |
| UpDn+SCRExpl.=HAT2019.05 | 62.2 | 78.9 | 41.4 | 54.3 | — | |
| UpDn+SCRExpl.=VQA-X2019.05 | 62.2 | 78.8 | 41.6 | 54.5 | — | |
| SAN-CTItraining_type=student model2019.09 | 62.1 | — | — | — | — | |
| SAN2019.09 | 61.7 | — | — | — | — | |
| fPMC2021.10 | 61.7 | — | — | — | — | |
| RUBi2019.06 | 61.16 | — | — | — | — | |
| VQ2AEnd-to-End Training=true, Shot Number=0, Access to Answer Candidates=true2022.12 | 61.1 | — | — | — | — | |
| UpDn+AttAlign2019.05 | 61 | 78.9 | 38.4 | 53.3 | — | |
| Img2LLM 175BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false, LLM Size=175B2022.12 | 60.6 | — | — | — | — | |
| Img2LLM 66BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false, LLM Size=66B2022.12 | 59.9 | — | — | — | — | |
| PICa-FullImage Repr.=GT-Caption-5, Few-shot=true, Oracle Performance=true2021.09 | 59.7 | — | — | — | — | |
| Show, Ask, Attend, and Answertraining_split=train2017.04 | 59.67 | 77.45 | 38.46 | 51.76 | — | |
| Img2LLM 30BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false, LLM Size=30B2022.12 | 59.5 | — | — | — | — | |
| MCBtraining_split=train2017.04 | 59.14 | 77.37 | 36.66 | 51.23 | — | |
| Img2LLM 6.7BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false, LLM Size=6.7B2022.12 | 57.6 | — | — | — | — | |
| Img2LLM 13BEnd-to-End Training=false, Shot Number=0, Access to Answer Candidates=false, LLM Size=13B2022.12 | 57.1 | — | — | — | — | |
| Flamingo 80BEnd-to-End Training=true, Shot Number=0, Access to Answer Candidates=false2022.12 | 56.3 | — | — | — | — | |
| PICa-FullImage Repr.=Caption+Tags, Few-shot=true2021.09 | 56.1 | — | — | — | — | |
| PICa175B-EnsembleEnd-to-End Training=false, Shot Number=80, Access to Answer Candidates=false2022.12 | 56.1 | — | — | — | — | |
| PICa-FullModel Size=175B, ND (Number of samples)=642023.03 | 56.1 | — | — | — | — | |
| PICa-FullImage Repr.=Caption, Few-shot=true2021.09 | 55.9 | — | — | — | — | |
| HieCoAtttraining_split=train2017.04 | 54.57 | 71.8 | 36.53 | 46.25 | — | |
| PICa-BaseImage Repr.=Caption+Tags, Few-shot=true2021.09 | 54.3 | — | — | — | — | |
| PICa175B+End-to-End Training=false, Shot Number=16, Access to Answer Candidates=false, Notes=Assumes access to training samples2022.12 | 54.3 | — | — | — | — | |
| PICa-BaseModel Size=175B, ND (Number of samples)=642023.03 | 54.3 | — | — | — | — | |
| PICa-BaseImage Repr.=Caption, Few-shot=true2021.09 | 53.2 | — | — | — | — | |
| FewVLM largeEnd-to-End Training=true, Shot Number=16, Access to Answer Candidates=false2022.12 | 51.1 | — | — | — | — | |
| TAP-CBackbone=Res50x16, k=32, Vdemo=true2022.03 | 50.18 | 73.51 | 31.56 | 37.35 | — | |
| MixPHMModel Size=226M, #Param (%)=0.39%, ND (Number of samples)=64, Tuning Mode=Parameter-efficient tuning of FewVLM2023.03 | 49.3 | — | — | — | — | |
| TAP-CBackbone=ViT-B/16, k=32, Vdemo=true2022.03 | 49.19 | 73.6 | 32.55 | 35.02 | — | |
| TAP-CBackbone=Res50x16, k=16, Vdemo=true2022.03 | 48.89 | 72.98 | 29.96 | 35.58 | — |