HyMem: Hierarchical Context Management for Long-Horizon Agents via Information Isolation

2026-08-16

AI Summary

This study addresses planning information loss and reasoning degradation in LLM agents during long-horizon tasks caused by context redundancy. We propose a hierarchical context management framework that functionally isolates planning and execution layers, employing independent reasoning modules to prevent subtask contamination of global memory while utilizing structured summarization to maintain long-range coherence. Experimental results demonstrate that this framework achieves Pass@1 scores of 66.7% on GAIA and 61.3% on BrowseComp-Plus, outperforming the strongest baselines by 6.1 and 4.7 percentage points, respectively. These findings confirm that our approach effectively mitigates context bloat and significantly enhances performance on complex, long-horizon tasks.

Abstract

Large language model (LLM) agents often perform poorly on complex, long-horizon tasks because their context becomes increasingly cluttered over time. As interactions accumulate, detailed execution traces and intermediate outputs dominate the context, making it difficult for the model to retain and use high-level planning information. Most existing methods address this issue through compression or retrieval applied to a single, flat context, which does not clearly separate different types of context information and often leads to degraded reasoning. To address this challenge, we propose HyMem, a hierarchical framework that explicitly separates the agent's context into distinct functional layers. HyMem organizes context by function to separate high-level planning from execution and complex analysis. Its isolated reasoning module handles complex subtasks without adding intermediate reasoning traces to the persistent planning context, while its memory management module preserves task progress across context refreshes through structured summaries. These components reduce redundant context accumulation, retain task-critical information, and support coherent long-horizon reasoning within a limited context window. Experiments on GAIA and Browsecomp-plus show that, with DeepSeek-V4, HyMem achieves average Pass@1 scores of 66.7% and 61.3%, outperforming the strongest baseline by 6.1 and 4.7 percentage points, respectively. Further analysis indicates that HyMem effectively controls the growth of the reasoning context, allowing the model to maintain focus and accuracy across complex, long-horizon tasks.

Problem

• Long-Horizon Agents
• Context Management
• Large Language Models
• Information Isolation
• Reasoning Degradation

Innovation

• Hierarchical Context Management
• Information Isolation
• Long-Horizon Agents
• Memory Management
• Context Window Optimization