·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Pruning LLMs Like a Physicist: Block Removal as an Ising Optimization Problem21m◆iPhone owners can now submit a claim in Apple’s $250 million Siri AI lawsuit28m◆The US spent billions on border surveillance. Why can’t it catch people before they die?2h◆How we made the first comprehensive map of deaths along the US border’s “virtual wall”2h◆4 ways to address the failures we found along the US border’s “virtual wall”2h◆She died at the San Diego border. A surveillance camera was in plain sight2h◆UN says AI safeguards can’t wait for certainty3h◆Amazon blocks Meta’s Muse AI agent4h◆LoRA Enhanced Contrastive Learning with SAS Vision Transformers10h◆Hallucination-R1: Robustness-Oriented Paraphrase Generation for Factual Consistency10h◆TatBLiMP: A Benchmark of Linguistic Minimal Pairs for Tatar10h◆Runtime Authorization for Resources Acquired by AI Agents10h◆JudgeSense: A Benchmark for Prompt Sensitivity in LLM-as-a-Judge Systems10h◆Multi-Domain Clustering via Measure Quantization10h◆Generative inversion for early ranking of competing geologic interpretations10h◆Do small language models know what they don't know?10h◆Listen Before You Speak: Response Planning from Listener Facial Reactions for Conversational Speech Generation10h◆On Repulsive and Attractive Teachers: Separating Correctness from Behavior in Self-Distillation10h◆Chinese Competitive Debating Dataset and Benchmark10h◆Intent-Governed Tool Authorization for AI Agents10h◆Pruning LLMs Like a Physicist: Block Removal as an Ising Optimization Problem21m◆iPhone owners can now submit a claim in Apple’s $250 million Siri AI lawsuit28m◆The US spent billions on border surveillance. Why can’t it catch people before they die?2h◆How we made the first comprehensive map of deaths along the US border’s “virtual wall”2h◆4 ways to address the failures we found along the US border’s “virtual wall”2h◆She died at the San Diego border. A surveillance camera was in plain sight2h◆UN says AI safeguards can’t wait for certainty3h◆Amazon blocks Meta’s Muse AI agent4h◆LoRA Enhanced Contrastive Learning with SAS Vision Transformers10h◆Hallucination-R1: Robustness-Oriented Paraphrase Generation for Factual Consistency10h◆TatBLiMP: A Benchmark of Linguistic Minimal Pairs for Tatar10h◆Runtime Authorization for Resources Acquired by AI Agents10h◆JudgeSense: A Benchmark for Prompt Sensitivity in LLM-as-a-Judge Systems10h◆Multi-Domain Clustering via Measure Quantization10h◆Generative inversion for early ranking of competing geologic interpretations10h◆Do small language models know what they don't know?10h◆Listen Before You Speak: Response Planning from Listener Facial Reactions for Conversational Speech Generation10h◆On Repulsive and Attractive Teachers: Separating Correctness from Behavior in Self-Distillation10h◆Chinese Competitive Debating Dataset and Benchmark10h◆Intent-Governed Tool Authorization for AI Agents10h◆
News/On Repulsive and Attractive Teachers: Separating Correctness from Behavior in Self-Distillation
arxiv
PublishedSeptember 21, 2026 at 4:00 AM
—neutral

On Repulsive and Attractive Teachers: Separating Correctness from Behavior in Self-Distillation

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

arXiv:2609.21561v1 Announce Type: cross Abstract: On-policy self-distillation provides dense, token-level supervision by conditioning a model on privileged information and distilling the resulting teacher distribution back into the model. However, privileged information can change not only what the

Stay posted· Newsletter

A 5-min weekly brief — top movers, price watch, story of the week.

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

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
arxivLoRA Enhanced Contrastive Learning with SAS Vision Transformers10harxivHallucination-R1: Robustness-Oriented Paraphrase Generation for Factual Consistency10harxivTatBLiMP: A Benchmark of Linguistic Minimal Pairs for Tatar10harxivRuntime Authorization for Resources Acquired by AI Agents10h
The Bubble Brief
WEEKLY

Read AI insights every Tuesday — top movers, new releases, story of the week.

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

Originally published on arxiv ↗
HomeModelsNews