Cognitive Scaffold: From Fluid Context to Crystallized Memory for Long-Horizon DeepResearch Agents

作者：Association for Computational Linguistics 2026, Qiuyuan Ai, Zenghuang Fu, Guannan He, Ping Jiang, Z Li, Jie Song, Haoyu Wu · 发表于：Underline Science Inc. · 年份：2026 · DOI：10.48448/n74x-ze67 · 研究领域：Computer science、Artificial intelligence、Cognitive science、Theoretical computer science、Cognitive psychology、Machine learning、Human–computer interaction

Scaling LLM-based agents to long-horizon deep research is constrained by the context-noise trade-off, where linear history accumulation degrades reasoning and dilutes fine-grained evidence. To address this, we introduce the Cognitive Scaffold, a factorized memory architecture that decouples the cognitive state into a Fluid Working Context for immediate reasoning and a persistent Knowledge Graph for long-term retention. Unlike unstructured summarization, our framework employs a Rejection Sampling Fine-Tuning (RFT) pipeline to crystallize saturated context into structured event snapshots, strictly enforcing atomic constraints to preserve numerical values and entities. During reasoning, a thought-driven dual-path retrieval mechanism enables the agent to proactively recover precise evidence. Empirical evaluations on Xbench-DeepSearch, BrowseComp-ZH, and GAIA demonstrate that Cognitive Scaffold consistently outperforms baselines, achieving 74.7% Avg@3 and 87.0% Pass@3 on Xbench-DeepSearch, 48.5% Avg@3 and 65.9% Pass@3 on BrowseComp-ZH, and 72.8% Avg@3 and 88.3% Pass@3 on GAIA, while reducing compression hallucinations to 5.3%. We open-source our codebase to facilitate future research.