DAPO

Uses

How it works

Comparisons

Highlighted claims

• DAPO removes zero-variance groups after rollouts are complete. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
• DAPO saves training cost but does not save rollout cost. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
• Selective rollout differs from DAPO because it uses information revealed during an agent rollout. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL
• The article recommends combining selective rollout with DAPO in production. — Selective Rollout: Mid-Trajectory Termination for Multi-Sample Agent RL