·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
AI’s most important protocol is getting a little bit easier to use40m◆X relaunches a rebuilt Android app after year-long effort1h◆OpenAI is scared of open-weight models. Should the US be?1h◆China’s AI models have Trump’s AI world at war with itself3h◆Adobe’s ‘natural look’ camera app embraces generative AI5h◆Introducing Cosmos 3 Edge5h◆Adobe camera app’s new feature will critique your photos using AI5h◆YouTube clarifies policies around AI slop and upsetting videos6h◆China delivers a one-two punch to America’s AI dominance11h◆Safety and alignment in an era of long-horizon models11h◆AI is more likely than humans to form biases when hiring12h◆ADS-C: Antidistillation Sampling for Classification17h◆Beyond a Joke: Multi-Angle Reasoning for Detecting and Explaining Harmful Humor in Memes17h◆From Black Box to Executable Logic: Explainable Reinforcement Learning through Prolog Expert Systems17h◆A Formally Grounded ODRL Evaluator: Implementation and Comparison17h◆Efficient Difficulty-Aware Dynamic Routing for Diffusion-Based Real-World Image Super-Resolution17h◆Digital Pantheon: Simulating and Auditing Coalition Formation with LLM Agents17h◆EpiNarrate: Agentic Generation of Grounded Narratives from Epidemiological Scenario Projections17h◆Before the Action: Benchmarking LLMs on Prospective Hypothesis Discovery17h◆How Much Human Label Variation Does Formal Semantic Structure Explain?: Group-Level Effects and Item-Level Ceilings in NLI17h◆AI’s most important protocol is getting a little bit easier to use40m◆X relaunches a rebuilt Android app after year-long effort1h◆OpenAI is scared of open-weight models. Should the US be?1h◆China’s AI models have Trump’s AI world at war with itself3h◆Adobe’s ‘natural look’ camera app embraces generative AI5h◆Introducing Cosmos 3 Edge5h◆Adobe camera app’s new feature will critique your photos using AI5h◆YouTube clarifies policies around AI slop and upsetting videos6h◆China delivers a one-two punch to America’s AI dominance11h◆Safety and alignment in an era of long-horizon models11h◆AI is more likely than humans to form biases when hiring12h◆ADS-C: Antidistillation Sampling for Classification17h◆Beyond a Joke: Multi-Angle Reasoning for Detecting and Explaining Harmful Humor in Memes17h◆From Black Box to Executable Logic: Explainable Reinforcement Learning through Prolog Expert Systems17h◆A Formally Grounded ODRL Evaluator: Implementation and Comparison17h◆Efficient Difficulty-Aware Dynamic Routing for Diffusion-Based Real-World Image Super-Resolution17h◆Digital Pantheon: Simulating and Auditing Coalition Formation with LLM Agents17h◆EpiNarrate: Agentic Generation of Grounded Narratives from Epidemiological Scenario Projections17h◆Before the Action: Benchmarking LLMs on Prospective Hypothesis Discovery17h◆How Much Human Label Variation Does Formal Semantic Structure Explain?: Group-Level Effects and Item-Level Ceilings in NLI17h◆
News/Nested-ReFT: Efficient Reinforcement Learning for Large Language Model Fine-Tuning via Off-Policy Rollouts
arxiv
PublishedJuly 14, 2026 at 4:00 AM

Nested-ReFT: Efficient Reinforcement Learning for Large Language Model Fine-Tuning via Off-Policy Rollouts

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

arXiv:2508.10123v3 Announce Type: replace-cross Abstract: Advanced reasoning in LLMs on challenging domains like mathematical reasoning can be tackled using verifiable rewards based reinforced fine-tuning (ReFT). In standard ReFT frameworks, a behavior model generates multiple completions with answe

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 Classification17harxivBeyond a Joke: Multi-Angle Reasoning for Detecting and Explaining Harmful Humor in Memes17harxivFrom Black Box to Executable Logic: Explainable Reinforcement Learning through Prolog Expert Systems17harxivA Formally Grounded ODRL Evaluator: Implementation and Comparison17h
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