Skip to content

Add std_thresholdoption to StepWiseGRPOAdvantageFn, to filter out zero-grad group samples.#363

Merged
pan-x-c merged 5 commits intoagentscope-ai:mainfrom
garyzhang99:feature/dynamic_sampling_for_multistep_grpo
Nov 3, 2025
Merged

Add std_thresholdoption to StepWiseGRPOAdvantageFn, to filter out zero-grad group samples.#363
pan-x-c merged 5 commits intoagentscope-ai:mainfrom
garyzhang99:feature/dynamic_sampling_for_multistep_grpo

Commits

Commits on Nov 3, 2025

Comments