·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Trump’s latest AI czar has already resigned1h◆Here are the 30,000 songs Sony is suing Udio’s AI music generator over1h◆Google is working on a new AI chip designed to make Gemini more efficient1h◆SpaceX in your index fund, explained2h◆AI’s most important protocol is getting a little bit easier to use2h◆X relaunches a rebuilt Android app after year-long effort3h◆OpenAI is scared of open-weight models. Should the US be?3h◆China’s AI models have Trump’s AI world at war with itself5h◆Adobe’s ‘natural look’ camera app embraces generative AI7h◆Introducing Cosmos 3 Edge7h◆Adobe camera app’s new feature will critique your photos using AI7h◆YouTube clarifies policies around AI slop and upsetting videos7h◆China delivers a one-two punch to America’s AI dominance13h◆Safety and alignment in an era of long-horizon models13h◆AI is more likely than humans to form biases when hiring14h◆ADS-C: Antidistillation Sampling for Classification19h◆Beyond a Joke: Multi-Angle Reasoning for Detecting and Explaining Harmful Humor in Memes19h◆From Black Box to Executable Logic: Explainable Reinforcement Learning through Prolog Expert Systems19h◆A Formally Grounded ODRL Evaluator: Implementation and Comparison19h◆Efficient Difficulty-Aware Dynamic Routing for Diffusion-Based Real-World Image Super-Resolution19h◆Trump’s latest AI czar has already resigned1h◆Here are the 30,000 songs Sony is suing Udio’s AI music generator over1h◆Google is working on a new AI chip designed to make Gemini more efficient1h◆SpaceX in your index fund, explained2h◆AI’s most important protocol is getting a little bit easier to use2h◆X relaunches a rebuilt Android app after year-long effort3h◆OpenAI is scared of open-weight models. Should the US be?3h◆China’s AI models have Trump’s AI world at war with itself5h◆Adobe’s ‘natural look’ camera app embraces generative AI7h◆Introducing Cosmos 3 Edge7h◆Adobe camera app’s new feature will critique your photos using AI7h◆YouTube clarifies policies around AI slop and upsetting videos7h◆China delivers a one-two punch to America’s AI dominance13h◆Safety and alignment in an era of long-horizon models13h◆AI is more likely than humans to form biases when hiring14h◆ADS-C: Antidistillation Sampling for Classification19h◆Beyond a Joke: Multi-Angle Reasoning for Detecting and Explaining Harmful Humor in Memes19h◆From Black Box to Executable Logic: Explainable Reinforcement Learning through Prolog Expert Systems19h◆A Formally Grounded ODRL Evaluator: Implementation and Comparison19h◆Efficient Difficulty-Aware Dynamic Routing for Diffusion-Based Real-World Image Super-Resolution19h◆
News/Fragile Knowledge, Robust Instruction-Following: The Width Pruning Dichotomy in Llama-3.2
arxiv
PublishedJune 15, 2026 at 4:00 AM
—neutral

Fragile Knowledge, Robust Instruction-Following: The Width Pruning Dichotomy in Llama-3.2

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

arXiv:2512.22671v3 Announce Type: replace-cross Abstract: Structured width pruning of GLU-MLP layers in Llama-3.2 models, guided by the Peak-to-Peak Magnitude (PPM) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities. While performance o

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
arxivADS-C: Antidistillation Sampling for Classification19harxivBeyond a Joke: Multi-Angle Reasoning for Detecting and Explaining Harmful Humor in Memes19harxivFrom Black Box to Executable Logic: Explainable Reinforcement Learning through Prolog Expert Systems19harxivA Formally Grounded ODRL Evaluator: Implementation and Comparison19h
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