·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
Making the case for context-aware AI1h◆AI’s hidden storage bill: What inference leaves behind3h◆Ensemble forecasting with IBM Granite time-series models4h◆Retrieval Augmented Generation Compositionality7h◆Evolving with Open Worlds: Towards Recursive Self-Improvement for Future Forecasting7h◆Towards English - Mizo Legal Machine Translation: Corpus Creation, Analysis, and Benchmarks7h◆VeriEvolve: Evaluator-First Evolution for Tool-Using Agents7h◆Beyond Model Weights: Sampling-Dependent Fairness in Diffusion Language Models7h◆Asymmetry Tunneling Beyond Adaptive Equivariance Relaxation7h◆Asymmetric Likelihood-Free Hypothesis Testing7h◆Bilateral Sharpness-Aware Minimization for Flatter Minima7h◆Backtracking Direct, Indirect and Total Effects7h◆TNODEV: Toolbox for Neural ODE Verification7h◆How to Use Parallel Data for Multilingual Pretraining7h◆KIRA: A Full-Stack Reading Screening System with Dual-Pathway Analysis for Children7h◆Building AI Research Teams for Autonomous Discovery of Linear Attention Architectures7h◆CSD: Common-Support Distillation for Cross-Tokenizer Knowledge Transfer7h◆Is Refusal Direction Universal across Low-Resourced Languages?7h◆Guilt by Association: Semantic Leakage Drives Associative Gender Bias7h◆Scan or Index? large Evidence Coverage, Answer Quality, and Cost in Long-Document RAG7h◆Making the case for context-aware AI1h◆AI’s hidden storage bill: What inference leaves behind3h◆Ensemble forecasting with IBM Granite time-series models4h◆Retrieval Augmented Generation Compositionality7h◆Evolving with Open Worlds: Towards Recursive Self-Improvement for Future Forecasting7h◆Towards English - Mizo Legal Machine Translation: Corpus Creation, Analysis, and Benchmarks7h◆VeriEvolve: Evaluator-First Evolution for Tool-Using Agents7h◆Beyond Model Weights: Sampling-Dependent Fairness in Diffusion Language Models7h◆Asymmetry Tunneling Beyond Adaptive Equivariance Relaxation7h◆Asymmetric Likelihood-Free Hypothesis Testing7h◆Bilateral Sharpness-Aware Minimization for Flatter Minima7h◆Backtracking Direct, Indirect and Total Effects7h◆TNODEV: Toolbox for Neural ODE Verification7h◆How to Use Parallel Data for Multilingual Pretraining7h◆KIRA: A Full-Stack Reading Screening System with Dual-Pathway Analysis for Children7h◆Building AI Research Teams for Autonomous Discovery of Linear Attention Architectures7h◆CSD: Common-Support Distillation for Cross-Tokenizer Knowledge Transfer7h◆Is Refusal Direction Universal across Low-Resourced Languages?7h◆Guilt by Association: Semantic Leakage Drives Associative Gender Bias7h◆Scan or Index? large Evidence Coverage, Answer Quality, and Cost in Long-Document RAG7h◆
News/How Much Gradient Does Truncated Backprop Lose in Looped Transformers?
arxiv
PublishedOctober 22, 2026 at 4:00 AM

How Much Gradient Does Truncated Backprop Lose in Looped Transformers?

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

arXiv:2610.28871v1 Announce Type: new Abstract: Looped transformers reuse one block of layers many times and are often trained by backpropagating through only the last k loops. We measure how much of the weight gradient this throws away, exactly and for every k from one backward pass, on nine looped

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
arxivRetrieval Augmented Generation Compositionality7harxivEvolving with Open Worlds: Towards Recursive Self-Improvement for Future Forecasting7harxivTowards English - Mizo Legal Machine Translation: Corpus Creation, Analysis, and Benchmarks7harxivVeriEvolve: Evaluator-First Evolution for Tool-Using Agents7h
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