CompactionRL: Reinforcement Learning with Context Compaction for Long-Horizon Agents

Zhipu AI / Tsinghua University

Research official 1 src. ~1 min

CompactionRL addresses context-window limitations in long-horizon agentic RL by jointly training task execution and trajectory summarization. Using token-level loss normalization and cross-trajectory advantage estimation, it achieves 66.8% Pass@1 on SWE-Bench Verified (+7.0pp). The method was subsequently deployed in training GLM-5.2.

Why it matters

Demonstrates a practical path to scaling agentic coding models beyond context window constraints; the SWE-Bench improvement is meaningful and GLM-5.2 production deployment confirms real-world viability.

Importance: 3/5

Deployed in production (GLM-5.2 training); +7pp SWE-Bench Verified improvement for long-horizon coding agents.

rl long-context coding agentic reasoning swe-bench glm

Sources

official arXiv:2607.05378 — CompactionRL