·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics54m◆A Consensus-Based Framework for Relative Preference Evaluation of Large Language Models6h◆Probing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders6h◆Data Quality over Capacity: Internalizing Documents into LoRA Adapters for Closed-Book QA6h◆Enjoy Your Talk: A Human-Centered Benchmark for Multi-Turn Dialogue with Decoupled User Simulation, Target Modeling, and Judging6h◆Multi-Mask Diffusion Language Models for Few-Step Generation6h◆Solar Open 2 Technical Report6h◆The Geometry of Personality: Activation Steering with Jungian Cognitive Functions6h◆Self-Guided Process Reward Optimization with Redefined Step-wise Advantage for Process Reinforcement Learning6h◆H$^2$SD: Hybrid Hindsight Self-Distillation6h◆LunarFM: A Shared Multimodal Representation of the Moon's Surface6h◆Prior laundering: learned priors with inherited, undetectable overconfidence6h◆Deep Sigma Point Processes for RCS Modeling in Spaceborne SAR Imagery6h◆Prompt as a Data Type: In-Database LLM Prompt Management and Rewriting6h◆CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference6h◆Quantum Spectral Model: Data Reuploading with Input-Conditioned Frequency Support6h◆Meta-Learning Approaches for Speaker-Dependent Voice Fatigue Models6h◆A Comparative Benchmark of Federated Learning Strategies for Mortality Prediction on Heterogeneous and Imbalanced Clinical Data6h◆Simpson's Paradox in Behavioral Curves: How Aggregation Distorts Parametric Models of User Dynamics6h◆Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO6h◆NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics54m◆A Consensus-Based Framework for Relative Preference Evaluation of Large Language Models6h◆Probing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders6h◆Data Quality over Capacity: Internalizing Documents into LoRA Adapters for Closed-Book QA6h◆Enjoy Your Talk: A Human-Centered Benchmark for Multi-Turn Dialogue with Decoupled User Simulation, Target Modeling, and Judging6h◆Multi-Mask Diffusion Language Models for Few-Step Generation6h◆Solar Open 2 Technical Report6h◆The Geometry of Personality: Activation Steering with Jungian Cognitive Functions6h◆Self-Guided Process Reward Optimization with Redefined Step-wise Advantage for Process Reinforcement Learning6h◆H$^2$SD: Hybrid Hindsight Self-Distillation6h◆LunarFM: A Shared Multimodal Representation of the Moon's Surface6h◆Prior laundering: learned priors with inherited, undetectable overconfidence6h◆Deep Sigma Point Processes for RCS Modeling in Spaceborne SAR Imagery6h◆Prompt as a Data Type: In-Database LLM Prompt Management and Rewriting6h◆CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference6h◆Quantum Spectral Model: Data Reuploading with Input-Conditioned Frequency Support6h◆Meta-Learning Approaches for Speaker-Dependent Voice Fatigue Models6h◆A Comparative Benchmark of Federated Learning Strategies for Mortality Prediction on Heterogeneous and Imbalanced Clinical Data6h◆Simpson's Paradox in Behavioral Curves: How Aggregation Distorts Parametric Models of User Dynamics6h◆Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO6h◆
News/Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO
arxiv
PublishedJuly 27, 2026 at 4:00 AM
—neutral

Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO

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

arXiv:2605.30789v3 Announce Type: replace Abstract: We identify a new dimension for enhancing rollout diversity in Group Relative Policy Optimization (GRPO) for LLMs. While GRPO relies on diverse rollouts, prevailing strategies primarily increase diversity by injecting more token-level randomness, w

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
arxivA Consensus-Based Framework for Relative Preference Evaluation of Large Language Models6harxivProbing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders6harxivData Quality over Capacity: Internalizing Documents into LoRA Adapters for Closed-Book QA6harxivEnjoy Your Talk: A Human-Centered Benchmark for Multi-Turn Dialogue with Decoupled User Simulation, Target Modeling, and Judging6h
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