arXiv - CSCL: "Don't throw away your value model! Making PPO eve…" - Qoto Mastodon

arXiv - CSCL @arxiv_cscl@qoto.org

Don't throw away your value model! Making PPO even better via Value-Guided Monte-Carlo Tree Search decoding. (arXiv:2309.15028v2 [cs.CL] UPDATED)

http://arxiv.org/abs/2309.15028 #arXiv #NLProc

Oct 19, 2023, 03:18 · · arxiv-cscl · · ·

Sign in to participate in the conversation