β-OPSD: Deriving with Policy Optimization, Training with Self-Distillation
On-policy self-distillation (OPSD) is a promising approach to improve reasoning language models, but it remains brittle in practice: making it work reliably often requires substantial engineering effort. We identify a structural source of this difficulty: vanilla OPSD is precisely the β=1 member of a broader policy-optimization family, where β weights the KL penalty anchoring the student to a reference policy. This equivalence turns β from an implicit value fixed at one into a controllable regul
Record details
Published: 29 July 2026
Source: HuggingFace Daily Papers
Category: Research
Topics: Regulation · Children & education
Retrieved: 1 August 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.
$β$-OPSD: Deriving with Policy Optimization, Training with Self-Distillation
arXiv cs.LG · 30 July 2026
VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation
HuggingFace Daily Papers · 29 July 2026
Flux-OPD: On-Policy Distillation with Evolving Contexts
HuggingFace Daily Papers · 29 July 2026
Creative Task Cards for Reflection, Self-Efficiency and Self-Regulation in CS1 Introductory Programming: Initial Insights
arXiv cs.HC · 30 July 2026
From Scoring to Acting: Outcome-Verified Comparative Self-Distillation for LLM Agents
arXiv · 30 July 2026
Technology-Enhanced Tabletop Exercises for Cybersecurity Education: Lessons Learned
arXiv · 30 July 2026
How to cite this record
ethics.ai (29 July 2026), “β-OPSD: Deriving with Policy Optimization, Training with Self-Distillation,” evidence record 15305, https://ethics.ai/record/15305 (originally published by HuggingFace Daily Papers).
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.