·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study2h◆Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?2h◆Encode Once, Answer Many: Internalizing Context via Parametric Flow2h◆Masked Erosion: How Acceptance Gates Lose Redundant Safeguards in Self-Improving Agents2h◆Predict the Reward, Not the World: World Reward Models for ML Agents2h◆DES-JEPA: Learning a Shared Vocabulary for High-Dimensional Discrete Event Sequences2h◆Accelerating Nonlinear MPC by Behavior Cloning with Separated Timescales: Application to Nuclear Reactor Load-Following2h◆Value Probes: Learned Value Functions for Beam-Search Red Teaming2h◆From Leniency to Suspicion: How an Open Post-Training Recipe Changes Whom LLM Fact-Checkers Believe2h◆An Alignment Benchmark for Enterprise Agents on Era by Eon2h◆PT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation2h◆Right Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries2h◆"Close enough?" An investigation into multiplication errors in LLMs2h◆Curriculum-Based Deep Reinforcement Learning for Robust Autonomous Rocket Landing under Atmospheric Uncertainty2h◆Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models2h◆SocialArena: Evaluating Social Intelligence in Large Language Models2h◆ATLAS-Bench: An Agentic Tool-Use Benchmark for Local Area Search2h◆Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?2h◆Integrated Feature Selection for classification and Representation Learning with Group-Lasso Siamese Networks2h◆PAPERBENCHX: Can AI Reproduce Science Across Disciplines?2h◆Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study2h◆Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?2h◆Encode Once, Answer Many: Internalizing Context via Parametric Flow2h◆Masked Erosion: How Acceptance Gates Lose Redundant Safeguards in Self-Improving Agents2h◆Predict the Reward, Not the World: World Reward Models for ML Agents2h◆DES-JEPA: Learning a Shared Vocabulary for High-Dimensional Discrete Event Sequences2h◆Accelerating Nonlinear MPC by Behavior Cloning with Separated Timescales: Application to Nuclear Reactor Load-Following2h◆Value Probes: Learned Value Functions for Beam-Search Red Teaming2h◆From Leniency to Suspicion: How an Open Post-Training Recipe Changes Whom LLM Fact-Checkers Believe2h◆An Alignment Benchmark for Enterprise Agents on Era by Eon2h◆PT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation2h◆Right Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries2h◆"Close enough?" An investigation into multiplication errors in LLMs2h◆Curriculum-Based Deep Reinforcement Learning for Robust Autonomous Rocket Landing under Atmospheric Uncertainty2h◆Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models2h◆SocialArena: Evaluating Social Intelligence in Large Language Models2h◆ATLAS-Bench: An Agentic Tool-Use Benchmark for Local Area Search2h◆Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?2h◆Integrated Feature Selection for classification and Representation Learning with Group-Lasso Siamese Networks2h◆PAPERBENCHX: Can AI Reproduce Science Across Disciplines?2h◆
News/Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models
arxiv
PublishedOctober 31, 2026 at 4:00 AM

Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models

Source
arxiv.orgfull article ↗
Read on arxiv→
Publisher summary· verbatim

arXiv:2610.38681v1 Announce Type: new Abstract: Self-distillation offers a way to improve reasoning models when a suitable external teacher is unavailable. However, common approaches rely on privileged information to construct a stronger self-teacher, limiting their use when only unlabeled prompts a

Stay posted· Newsletter

Three short emails a week: Monday, Wednesday and Friday. Top movers, new models and the most-referenced story.

// no spam · unsubscribe one-click · free forever

Your email is used only to send The Bubble Brief. Privacy policy

Discussion
Source
↗
arxiv
Read original ↗All from arxiv →

No replies yet. Be first.

Source
↗
arxiv
Read original ↗All from arxiv →

Related coverage

More from ARXIV
arxivGenerating Critic-Guided Synthetic Experience in Latent World Models2harxivAppreciation Teaches Unauthorized Execution: Self-Evolution Degrades the Robustness of Personal Agents to Indirect Prompt Injection2harxivPRISM: Personalized Item Representations in User History2harxivRetroChessBench: A Benchmark for Reconstructing and Proving Chess Histories2h
The Bubble Brief
MON · WED · FRI

Read AI insights every Monday, Wednesday and Friday — top movers, new models, story of the week.

// no spam · unsubscribe one-click · free forever

Your email is used only to send The Bubble Brief. Privacy policy

Originally published on arxiv ↗
PrivacyTermsRefundsLegal notice|AboutMethodologyContact

© 2026 Databubble

Built by Marouane Gazouzi · Readyo, AI interview coach

HomeModelsNews