KV4-RL: Unlocking Faster Asynchronous Rollouts for RLVR with INT4 KV Cache Quantization

arXiv:2610.14448 · cs.LG, cs.AI · Submitted 2026-10-10 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-10-10

Updated: 2026-10-10

Code: https://github.com/deepseek-ai/DeepEP

Terminology

• KV cache — 1× in this paper · explained in 3 episodes
• Quantization — 1× in this paper · explained in 12 episodes

Sources

• BitDecoding: Unlocking Tensor Cores for Long-Context LLMs with Low-Bit KV Cache
• Gemma 3 Technical Report
• SAW-INT4: System-Aware 4-Bit KV-Cache Quantization for Real-World LLM Serving
• Kimi K2: Open Agentic Intelligence
• Kimi k1.5: Scaling Reinforcement Learning with LLMs
• StaleFlow: Staleness-Aware Data Management for Mitigating Data Skewness in Fully Disaggregated RL Post-Training
• QuRL: Efficient Reinforcement Learning with Quantized Rollout
• PM-KVQ: Progressive Mixed-precision KV Cache Quantization for Long-CoT LLMs
• Proximal Policy Optimization Algorithms
• Qwen3 Technical Report
• Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures
• Group Sequence Policy Optimization
• APRIL: Active Partial Rollouts in Reinforcement Learning to Tame Long-tail Generation

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks