Multicultural Visual Reasoning on MaRVL translated English version (test)
73.47AccuracyMETER
Evaluation results
| Method | Links | ||
|---|---|---|---|
| METERData/Steps=4M/-, #Param=352M2023.01 | 73.47 | 8.86 | |
| ALBEFData/Steps=14M/-, #Param=210M2023.01 | 73.17 | 9.37 | |
| X-VLMData/Steps=16M/-, #Param=216M2023.01 | 73.02 | 11.39 | |
| GIVLData/Steps=3.17M/900K, #Param=112M2023.01 | 72.7 | 7.17 | |
| GIVLData/Steps=3.17M/500K, #Param=112M2023.01 | 72.5 | 6.56 | |
| VinVLData/Steps=5.65M/2M, #Param=112M2023.01 | 72.48 | 8.55 | |
| GIVL w/o LIKMData/Steps=3.17M/500K, #Param=112M2023.01 | 72.11 | — | |
| GIVL w/ CLIPData/Steps=3.17M/500K, #Param=199M2023.01 | 71.78 | — | |
| GIVL-BData/Steps=3.17M/500K, #Param=112M2023.01 | 70.26 | — | |
| VinVL*Data/Steps=3.17M/500K, #Param=112M2023.01 | 69.66 | 8.27 | |
| GIVL w/o LIECData/Steps=3.17M/500K, #Param=112M2023.01 | 68.58 | — | |
| ViLBERTData/Steps=3.3M/-, #Param=274M2023.01 | 66.53 | 10.87 |