Spaces:
Running
Running
Method,Compute,PPL,Truth,ROUGE,MAPO | |
DPO,1,0.0768753815825175,0.3986066518668836,0.9565206583678539,0.8712100761297452 | |
DPO,1,0.5996286412275281,0.3602207419553394,0.6659869365178336,0.10860875787554469 | |
Grad Ascent,1,0.01276005409910086,0.8739653533368998,0.023760115576687335,0.557119299008513 | |
Grad Ascent,1,0.9153365174437643,0.5117768071328227,0.13019174875223205,0.2736769500895253 | |
IDK,1,0.9575847571359651,0.14495165859171177,0.5215967278097287,0.9513180970650936 | |
IDK,1,0.5081723023409522,0.7260250131902866,0.46407442478973215,0.9008803129332287 | |
KL,1,0.23284021819861755,0.7973023013038227,0.9713336423092905,0.674807833567933 | |
KL,1,0.04023188230471908,0.7071738714102987,0.4663170982373773,0.6658539062921722 | |