SAM: State-Adaptive Memory for Long-Horizon Reasoning Agent

2026-05-23

AI Summary

This work addresses the challenge of long-horizon agent reasoning, where critical information is sparsely distributed and dynamically shifts with the agent’s state, rendering conventional history truncation, compression, or retrieval methods ineffective. The authors propose the State-Adaptive Memory (SAM) framework, which explicitly models the alignment between memory access and the agent’s current state for the first time. SAM enables intent-driven, dynamic recall of historical trajectories by combining lightweight memory cues with original trajectory segments. Notably, it operates without requiring retraining of the backbone model and leverages a hybrid optimization strategy integrating expert supervision and reinforcement learning. Evaluated across multiple benchmarks—including BrowseComp, BrowseComp-ZH, WideSearch, and HLE—SAM significantly outperforms strong existing baselines.

Abstract

Long-horizon agentic reasoning requires large language models to act over long interaction histories containing thoughts, tool calls, observations, and partial conclusions. The challenge is not merely that these histories grow long, but that information needed for the current decision may be scattered across distant steps and only become relevant later. Existing approaches address this difficulty by truncating the interaction history, compressing it into shorter surrogates, or retrieving selected parts of it for reuse, but they do not explicitly model how access to past interaction should adapt to the agent's evolving state. We instead cast long-horizon reasoning as a problem of state-adaptive memory. To this end, we propose State-Adaptive Memory~(SAM), a standalone framework that consolidates ongoing interaction into compact memory cues while preserving raw trajectory pages for intent-driven recall. These cues are not treated as replacements for history; rather, they serve as lightweight handles that allow the agent to reconstruct temporally distant information according to its current needs, without retraining the underlying backbone. We further optimize the memory module through expert-guided supervision and reinforcement learning, aligning it with trajectory-level utility. Across BrowseComp, BrowseComp-ZH, WideSearch, and HLE, SAM consistently outperforms strong baselines over diverse agent backbones. Our results suggest that explicit memory modeling provides a simple and effective foundation for long-horizon agentic reasoning.

Problem

• long-horizon reasoning
• state-adaptive memory
• interaction history
• agentic reasoning
• memory modeling

Innovation

• State-Adaptive Memory
• Long-Horizon Reasoning
• Memory Cues
• Intent-Driven Recall
• Trajectory-Level Utility