LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches

arXiv:2610.06647 · cs.CL · Submitted 2026-10-05 · Read on arXiv

cs.CL

Submitted: 2026-10-05

Updated: 2026-10-06

Code: https://github.com/NVIDIA-NeMo/RL

Terminology

• Reinforcement Learning — 1× in this paper · explained in 4 episodes

Sources

• Qwen Technical Report
• Training Deep Nets with Sublinear Memory Cost
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• Flora: Low-Rank Adapters Are Secretly Gradient Compressors
• LoRA: Low-Rank Adaptation of Large Language Models
• Adam: A Method for Stochastic Optimization
• Let's Verify Step by Step
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• HybridFlow: A Flexible and Efficient RLHF Framework
• GaLore 2: Large-Scale LLM Pre-Training by Gradient Low-Rank Projection
• Polar: Agentic RL on Any Harness at Scale
• OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks
• ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents
• Offline Training of Language Model Agents with Functions as Learnable Weights
• GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection
• PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallel

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering