List your company

Positive Movie Review Generation on IMDB (test)

0.99RewardRRHF-OP-128
0.520960.642730.76450.88627Apr 11, 2023

Evaluation results

MethodLinks
RRHF-OP-128KL penalty=none2023.040.9932.081
RRHFSetting=BP2023.040.86132.083
RRHFSetting=B2023.040.79932.077
PPOKL penalty=none2023.040.79642.916
NLPOKL penalty=none2023.040.77741.035
RRHF-OP-128KL penalty=0.12023.040.63532.088
PPOKL penalty=0.12023.040.62635.049
NLPOKL penalty=0.12023.040.6234.816
SFTSetting=baseline2023.040.53935.472