·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
China’s AI models have Trump’s AI world at war with itself1h◆Adobe’s ‘natural look’ camera app embraces generative AI3h◆Introducing Cosmos 3 Edge3h◆YouTube clarifies policies around AI slop and upsetting videos3h◆China delivers a one-two punch to America’s AI dominance9h◆Safety and alignment in an era of long-horizon models9h◆AI is more likely than humans to form biases when hiring10h◆ADS-C: Antidistillation Sampling for Classification15h◆Digital Pantheon: Simulating and Auditing Coalition Formation with LLM Agents15h◆EpiNarrate: Agentic Generation of Grounded Narratives from Epidemiological Scenario Projections15h◆Before the Action: Benchmarking LLMs on Prospective Hypothesis Discovery15h◆How Much Human Label Variation Does Formal Semantic Structure Explain?: Group-Level Effects and Item-Level Ceilings in NLI15h◆Learn to Memorize: Scalable Continual Learning in Semiparametric Models with Mixture-of-Neighbors Induction Memory15h◆Echoes Across Vietnam's Highlands, Delta, and Coast: A Multilingual Corpus for Cham, Khmer, and Tay-Nung15h◆Stochastic Reset Pathfinding: Path-Level Regret for Cascading Bandits over Graph Paths15h◆Publicly-Verifiable Certificates for Statistical Algorithms15h◆Do Generative Models Keep Time? A Time-Aware Evaluation of Synthetic Sequential Tabular Data15h◆QUADS: Stabilizing NVFP4 Reinforcement Learning for MoE via QUantization-error Alignment across Dual Sides15h◆CLaC@FinMMEval 2026 Task 3: Sentiment-Augmented Deep Reinforcement Learning for Active Trading -- An Alpha-Reward Approach15h◆DELUGE: Towards Continental-Scale Daily Pluvial Flood Damage Prediction via Interpretable Conditioning on Foundation Model Embeddings15h◆China’s AI models have Trump’s AI world at war with itself1h◆Adobe’s ‘natural look’ camera app embraces generative AI3h◆Introducing Cosmos 3 Edge3h◆YouTube clarifies policies around AI slop and upsetting videos3h◆China delivers a one-two punch to America’s AI dominance9h◆Safety and alignment in an era of long-horizon models9h◆AI is more likely than humans to form biases when hiring10h◆ADS-C: Antidistillation Sampling for Classification15h◆Digital Pantheon: Simulating and Auditing Coalition Formation with LLM Agents15h◆EpiNarrate: Agentic Generation of Grounded Narratives from Epidemiological Scenario Projections15h◆Before the Action: Benchmarking LLMs on Prospective Hypothesis Discovery15h◆How Much Human Label Variation Does Formal Semantic Structure Explain?: Group-Level Effects and Item-Level Ceilings in NLI15h◆Learn to Memorize: Scalable Continual Learning in Semiparametric Models with Mixture-of-Neighbors Induction Memory15h◆Echoes Across Vietnam's Highlands, Delta, and Coast: A Multilingual Corpus for Cham, Khmer, and Tay-Nung15h◆Stochastic Reset Pathfinding: Path-Level Regret for Cascading Bandits over Graph Paths15h◆Publicly-Verifiable Certificates for Statistical Algorithms15h◆Do Generative Models Keep Time? A Time-Aware Evaluation of Synthetic Sequential Tabular Data15h◆QUADS: Stabilizing NVFP4 Reinforcement Learning for MoE via QUantization-error Alignment across Dual Sides15h◆CLaC@FinMMEval 2026 Task 3: Sentiment-Augmented Deep Reinforcement Learning for Active Trading -- An Alpha-Reward Approach15h◆DELUGE: Towards Continental-Scale Daily Pluvial Flood Damage Prediction via Interpretable Conditioning on Foundation Model Embeddings15h◆
News/Procedural-skill SFT across capacity tiers: A W-Shaped pre-SFT Trajectory and Regime-Asymmetric Mechanism on 0.8B-4B Qwen3.5 Models
arxiv
PublishedMay 15, 2026 at 4:00 AM
—neutral

Procedural-skill SFT across capacity tiers: A W-Shaped pre-SFT Trajectory and Regime-Asymmetric Mechanism on 0.8B-4B Qwen3.5 Models

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

arXiv:2605.11907v2 Announce Type: replace Abstract: We measure procedural-skill SFT contribution across three Qwen3.5 dense scales (0.8B, 2B, 4B) on a 200-task / 40-skill holdout, with Claude Haiku 4.5 as a frontier reference. The corpus is 353 rows of (task + procedural-skill block, Opus chain-of-t

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 Classification15harxivDigital Pantheon: Simulating and Auditing Coalition Formation with LLM Agents15harxivEpiNarrate: Agentic Generation of Grounded Narratives from Epidemiological Scenario Projections15harxivBefore the Action: Benchmarking LLMs on Prospective Hypothesis Discovery15h
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