CompactionRL: Reinforcement Learning with Context Compaction for Long-Horizon Agents
Long-horizon agentic LLMs are increasingly limited by finite context windows, as extended interaction trajectories can exceed the maximum context length before a task is completed. Context compaction offers a natural solution by summarizing previous interaction states and continuing the rollout under a compressed context, but incorporating compaction into reinforcement learning remains underexplored. We propose CompactionRL, a reinforcement learning strategy to train long-horizon agentic LLMs with context compaction. Our approach jointly optimizes task execution and summary generation with tok
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 51%RimantasZ/contextspy →
- PossiblePossibly related (embedding) · 49%AgentCore-8B →
- PossiblePossibly related (embedding) · 48%rllm-org/rllm →
- PossiblePossibly related (embedding) · 47%Context-Engine-AI/Context-Engine →
- PossiblePossibly related (embedding) · 46%redai-infra/Relax →
- FuzzySimilar title/name (fuzzy) · 87%NirDiamant/GenAI_Agents →
“Fuzzy title match (0.94): “CompactionRL: Reinforcement Learning with Context Compaction” ≈ “NirDiamant/GenAI_Agents””
- FuzzySimilar title/name (fuzzy) · 84%Thysrael/Horizon →
“Fuzzy title match (0.92): “CompactionRL: Reinforcement Learning with Context Compaction” ≈ “Thysrael/Horizon””
- FuzzySimilar title/name (fuzzy) · 84%Unity-Technologies/ml-agents →
“Fuzzy title match (0.92): “CompactionRL: Reinforcement Learning with Context Compaction” ≈ “Unity-Technologies/ml-agents””