rl-reward

SKILLFlujo de trabajocomunidad
v0.0.0agentscope-aiApache-2.0Actualizado hace 17 dFuente →

Build RL reward signals using the OpenJudge framework. Covers choosing between pointwise and pairwise reward strategies based on RL algorithm, task type, and cost; aggregating multi-dimensional pointwise scores into a scalar reward; pairwise tournament reward for GRPO on subjective tasks (net win ra

Community-submitted skill. Not yet reviewed by the Forge team. Full prompt content may not be available.Request review →
789Estrellas del repo
1Clientes
1Formatos
hace 17 dÚltima actualización
Skill
Autoragentscope-ai
Versión0.0.0
LicenciaApache-2.0
CategoríaFlujo de trabajo
Formatosskill.md
PromptAbrir (ver la pestaña Prompt)
Compatibilidad
Claude✓ Compatible
Cursor
Copilot
ChatGPT
Gemini
Acerca de

Build RL reward signals using the OpenJudge framework. Covers choosing between pointwise and pairwise reward strategies based on RL algorithm, task type, and cost; aggregating multi-dimensional pointwise scores into a scalar reward; pairwise tournament reward for GRPO on subjective tasks (net win rate across group rollouts); generating preference pairs for DPO/RLAIF; and normalizing scores for tra

Palabras clave
skillclaude