Skip to content

fix(kl): apply the reward KL penalty only when ref_log_prob is present - #639

Open
flyingdoubleG wants to merge 1 commit into
agentscope-ai:mainfrom
flyingdoubleG:fix/kl-penalty-without-ref
Open

fix(kl): apply the reward KL penalty only when ref_log_prob is present#639
flyingdoubleG wants to merge 1 commit into
agentscope-ai:mainfrom
flyingdoubleG:fix/kl-penalty-without-ref

Commits

  1. Commits on Oct 20, 2026