KV-streams for Efficient Compaction in Agentic Reinforcement Learning

arXiv:2609.35750 · cs.LG, cs.AI · Submitted 2026-09-28 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-09-28

Updated: 2026-10-15

Code: https://github.com/Emilianopp/KV-streams

Terminology

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

Sources

• Libra: Efficient Resource Management for Agentic RL Post-Training
• Parallel Context Compaction for Long-Horizon LLM Agent Serving
• TALES: Text Adventure Learning Environment Suite
• Sliding Window Attention Training for Efficient Large Language Models
• Sliding-window beats linear attention
• Self-Compacting Language Model Agents
• CompactionRL: Reinforcement Learning with Context Compaction for Long-Horizon Agents
• Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Management
• Prefix Sliding for efficient test-time scaling
• MemGPT: Towards LLMs as Operating Systems
• INTELLECT-3: Technical Report
• ScienceWorld: Is your Agent Smarter than a 5th Grader?
• EvoMemBench: Benchmarking Agent Memory from a Self-Evolving Perspective
• ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization
• Efficient Streaming Language Models with Attention Sinks
• InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement Learning
• Immersion in the GitHub Universe: Scaling Coding Agents to Mastery

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