Skip to content

fix(rewards): make DAPO zero-penalty region inclusive so cache_length=0 no longer divides by zero - #627

Open
Linxiushen wants to merge 1 commit into
agentscope-ai:mainfrom
Linxiushen:fix-dapo-overlong-zero-cache-length
Open

Linxiushen wants to merge 1 commit into
agentscope-ai:mainfrom
Linxiushen:fix-dapo-overlong-zero-cache-length

Commits

Commits on Sep 21, 2026