List your company

Multimodal Question Answering on ManyModalQA (test)

92.5Accuracy (Text)MAMMQA
20.94839.52458.176.676May 27, 2025

Evaluation results

MethodLinks
MAMMQABackbone=OpenAI 4o-mini2025.0592.596.7878.0289.9
Human2025.059289.69491.6
MAMMQABackbone=Gemini 1.5-Flash 8B2025.0589.7694.5277.3387.91
MAMMQABackbone=Qwen 2.5 VL 3B Instruct2025.0588.7994.972.6786.37
CapCoTBackbone=OpenAI 4o-mini2025.0587.6894.0568.2684.41
CoTBackbone=OpenAI 4o-mini, No-context variant=false2025.0587.294.2357.3381.21
MAMMQABackbone=Qwen 2.5 VL 7B Instruct2025.0587.1196.3177.5687.61
ToTBackbone=Gemini 1.5-Flash 8B2025.0586.0886.8162.8179.8
CoTBackbone=Gemini 1.5-Flash 8B, No-context variant=false2025.0586.0591.5268.7782.81
CapCoTBackbone=Gemini 1.5-Flash 8B2025.0585.7491.463.1481.34
ToTBackbone=OpenAI 4o-mini2025.0584.9493.1972.984.7
CapCoTBackbone=Qwen 2.5 VL 7B Instruct2025.0583.592.8671.0783.41
ToTBackbone=Qwen 2.5 VL 3B Instruct2025.0582.6686.1468.1180.42
ToTBackbone=Qwen 2.5 VL 7B Instruct2025.0581.9590.4169.2981.89
CapCoTBackbone=Qwen 2.5 VL 3B Instruct2025.0580.7991.3867.1380.63
CoTBackbone=Qwen 2.5 VL 3B Instruct, No-context variant=false2025.0570.0875.6150.766.54
CoTBackbone=OpenAI 4o-mini, No-context variant=true2025.0568.2270.5159.4266.54
CoTBackbone=Qwen 2.5 VL 7B Instruct, No-context variant=true2025.0561.866.7354.5361.46
CoTBackbone=Qwen 2.5 VL 7B Instruct, No-context variant=false2025.0559.8468.7145.4758.87
MMQAOracle setting=true2025.0559.346.32946.3
CoTBackbone=Qwen 2.5 VL 3B Instruct, No-context variant=true2025.0558.7764.5559.5158.77
CoTBackbone=Gemini 1.5-Flash 8B, No-context variant=true2025.0554.9361.1534.7751.41
UniMMQABackbone=T5 Model, Model Size=3B2025.0549.858.340.952.1
MMQAOracle setting=false2025.0548.640.427.239.7
UniMMQABackbone=T5 Model, Model Size=Large2025.0548.567.534.950
UniMMQABackbone=T5 Model, Model Size=Base2025.0546.660.730.245.4
Voting2025.0523.722.915.521.1