RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning¶

L1 core-mechanism diagnostic. Runs the retirement rule and student updates, not paper-scale agent-teacher training.

Paper information¶

Field	Details
Paper	RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning
Organization	First-author affiliation: Zhejiang University
First public release	2026-09-17 (arXiv v1)
Authors' code	ZJU-REAL/SDAR is public.
Adapter / method	`retire-opd`
Local reproduction code	`src/auto_research/post_training/latest_20260919.py`

Original paper¶

Motivation and method¶

Assign separate teachers to skills. The student combines RL and on-policy distillation, automatically retiring a teacher once success approaches the teacher's and the distribution gap stops narrowing.

flowchart LR
  I[Public input and current state] --> M[retire-opd mechanism]
  M --> A[Auditable intermediates]
  A --> O[Output or state update]

Key figure from the paper¶

Figure 3: Training pipeline and key optimization steps. From the original paper; copyright remains with the authors. Click the image for the source.

Core computation¶

The local reference kernel preserves the paper's decisive gates, weights, state transitions, or scheduling rules and records intermediates in the metrics artifact. Implementation functions and invariant assertions in tests map the formulas and variables.

Paper results¶

Reported online results, benchmarks, training efficiency, or inference speed are original paper results. The local three-seed artifact checks core mechanisms, shapes, and state invariants without comparing directly to the paper's numbers.

Reproduction boundaries¶

The retirement criterion and student update execute locally. The paper-scale agent teachers were not trained.