Skip to content
OpenTrain AIFor AI Companies
← Back to explorer

Blockwise Policy-Drift Gating for On-Policy Distillation

Liwen Zheng, Haiyun Jiang · Jun 23, 2026 · Citations: 0

How to use this page

Moderate trust

Use this for comparison and orientation, not as your only source.

Best use

Background context only

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Evidence quality

Moderate

Derived from extracted protocol signals and abstract evidence.

Abstract

On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories sampled by the student itself. Recent work shows that sampled-token OPD can be fragile on long-horizon reasoning tasks and that local teacher-support matching is a simple and effective repair. This paper introduces blockwise policy-drift gating, a lightweight student-only old-current drift controller for OPD under rollout reuse. The method computes log-probability shifts between the behavior student and the current student on the sampled token path, aggregates these shifts over fixed blocks or spans, and uses the resulting detached, mean-normalized gates to reweight OPD position losses. It does not change teacher targets, teacher top-K supports, or the rollout policy. In a six-variant Qwen3 math reasoning benchmark with a uniform 200-step training budget for all trained variants, we use pass@8 as the primary problem-level solve-rate metric. Fixed 64-token block gating improves sampled-token OPD mean pass@8 from 0.4978 to 0.5160 across AIME24, AIME25, MATH500, and AMC23. On Teacher-TopK/LSM, Block64 gives the best four-benchmark mean pass@8 among trained students. The results identify local old-current policy drift as a practical control signal for reused OPD rollouts and motivate block-level gating as a simple default for improving solve-rate robustness.

Should You Rely On This Paper?

This paper is adjacent to HFEPX scope and is best used for background context, not as a primary protocol reference.

Best use

Background context only

Use if you need

A benchmark-and-metrics comparison anchor.

Main weakness

No major weakness surfaced.

Trust level

Moderate

Usefulness score

25/100 • Low

Treat as adjacent context, not a core eval-method reference.

Human Feedback Signal

Not explicit in abstract metadata

Evaluation Signal

Detected

Usefulness for eval research

Adjacent candidate

Extraction confidence 55%

What We Could Verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

missing

None explicit

No explicit feedback protocol extracted.

"On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories sampled by the student itself."

Evaluation Modes

strong

Automatic Metrics

Includes extracted eval setup.

"On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories sampled by the student itself."

Quality Controls

missing

Not reported

No explicit QC controls found.

"On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories sampled by the student itself."

Benchmarks / Datasets

strong

MATH 500

Useful for quick benchmark comparison.

"On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories sampled by the student itself."

Reported Metrics

strong

Pass@8

Useful for evaluation criteria comparison.

"In a six-variant Qwen3 math reasoning benchmark with a uniform 200-step training budget for all trained variants, we use pass@8 as the primary problem-level solve-rate metric."

Human Feedback Details

  • Uses human feedback: No
  • Feedback types: None
  • Rater population: Not reported
  • Unit of annotation: Trajectory
  • Expertise required: Math

Evaluation Details

  • Evaluation modes: Automatic Metrics
  • Agentic eval: Long Horizon
  • Quality controls: Not reported
  • Evidence quality: Moderate
  • Use this page as: Background context only

Protocol And Measurement Signals

Benchmarks / Datasets

MATH-500

Reported Metrics

pass@8

Research Brief

Metadata summary

On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories sampled by the student itself.

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key Takeaways

  • On-policy distillation (OPD) trains a student policy using teacher signals computed on trajectories sampled by the student itself.
  • Recent work shows that sampled-token OPD can be fragile on long-horizon reasoning tasks and that local teacher-support matching is a simple and effective repair.
  • This paper introduces blockwise policy-drift gating, a lightweight student-only old-current drift controller for OPD under rollout reuse.

Researcher Actions

  • Compare this paper against others mentioning MATH.
  • Validate inferred eval signals (Automatic metrics, Long-horizon tasks) against the full paper.
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Recommended Queries

Research Summary

Contribution Summary

  • In a six-variant Qwen3 math reasoning benchmark with a uniform 200-step training budget for all trained variants, we use pass@8 as the primary problem-level solve-rate metric.
  • Fixed 64-token block gating improves sampled-token OPD mean pass@8 from 0.4978 to 0.5160 across AIME24, AIME25, MATH500, and AMC23.
  • On Teacher-TopK/LSM, Block64 gives the best four-benchmark mean pass@8 among trained students.

Why It Matters For Eval

  • In a six-variant Qwen3 math reasoning benchmark with a uniform 200-step training budget for all trained variants, we use pass@8 as the primary problem-level solve-rate metric.
  • On Teacher-TopK/LSM, Block64 gives the best four-benchmark mean pass@8 among trained students.

Researcher Checklist

  • Gap: Human feedback protocol is explicit

    No explicit human feedback protocol detected.

  • Pass: Evaluation mode is explicit

    Detected: Automatic Metrics

  • Gap: Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Pass: Benchmark or dataset anchors are present

    Detected: MATH-500

  • Pass: Metric reporting is present

    Detected: pass@8

Related Papers

Papers are ranked by protocol overlap, extraction signal alignment, and semantic proximity.