Less is More: Early Stopping Rollout for On-Policy Distillation
On-policy distillation has recently emerged as a promising alternative to standard sequence-level imitation, training a student by scoring its own rollouts with a teacher model. However, we observe ``Off-policy Teacher Decay'' problem in this paradigm: for the later tokens, with student's earlier trajectory as context that is off-policy to the teacher, the teacher's ability to produce a corrective score would decay, and may fall back to token-completion behavior learned in the pre-training stage
Record details
Published: 26 May 2026
Source: arXiv
Category: Research
Topics: Regulation · Children & education
Retrieved: 14 July 2026
Related evidence
These records share source-supplied organisations, an exact publisher byline, automatic topics or regions. The reason is shown on every link; related does not mean supporting, agreeing with or verifying this record.
Counteraction-Aware Multi-Teacher On-Policy Distillation for General Capability Recovery with Domain Preservation
arXiv · 26 May 2026
BenGER: Benchmarking LLM Systems on Subsumption-Based Legal Reasoning in German Law
arXiv · 27 May 2026
ADWIN: Adaptive Windows for Horizon-Aware On-Policy Distillation
arXiv · 27 May 2026
Generative AI as a Design Variable: An Evidence-Centered Framework for Principled Governance in STEM Assessment
arXiv · 24 May 2026
AI Sovereignty as National Learning Capacity: A Human-Centered Learning Mechanics Viewpoint on France, the United States, and China
arXiv · 30 May 2026
AVSD: Adaptive-View Self-Distillation by Balancing Consensus and Teacher-Specific Privileged Signals
arXiv · 20 May 2026
How to cite this record
ethics.ai (26 May 2026), “Less is More: Early Stopping Rollout for On-Policy Distillation,” evidence record 3676, https://ethics.ai/record/3676 (originally published by arXiv).
Use and limitations
This page is a stable index and citation surface for a source record. ethics.ai did not author the underlying report and has not independently verified every claim. Automatic topics may be imperfect. For consequential use, quote and cite the original publisher.