用强化学习训练小模型主动清理大模型的冗余信息，提升长任务成功率。
Escaping the Context Bottleneck: Active Context Curation for LLM Agents via Reinforcement Learning
- 用轻量级策略模型主动修剪上下文噪声，保留关键推理节点。
- 在WebArena上将成功率从36.4%提至41.2%，令牌消耗降8.8%。
- 70亿参数的小模型性能媲美GPT-4o，适合部署于资源受限场景。
大型语言模型在处理长周期任务时受限于‘上下文瓶颈’和‘中间迷失’现象，即冗长交互中累积的噪声会降低推理能力。为此，我们提出一种共生框架，将上下文管理与任务执行解耦。架构由轻量级专用策略模型ContextCurator与强大的冻结基础模型TaskExecutor组成。通过强化学习训练，ContextCurator主动降低工作记忆中的信息熵，激进地剔除环境噪声，同时保留对后续推断至关重要的稀疏关键数据点。在WebArena上，该框架将Gemini-3.0-flash的成功率从36.4%提升至41.2%，令牌消耗从47.4K降至43.3K（下降8.8%）。在DeepSearch上实现57.1%的成功率（原为53.9%），令牌消耗减少8倍。值得注意的是，仅70亿参数的ContextCurator即可达到GPT-4o级别的上下文管理性能，为自主长周期智能体提供可扩展且高效的计算范式。
原文摘要 · Abstract (English)
Large Language Models (LLMs) struggle with long-horizon tasks due to the "context bottleneck" and the "lost-in-the-middle" phenomenon, where accumulated noise from verbose environments degrades reasoning over multi-turn interactions. To address this issue, we introduce a symbiotic framework that decouples context management from task execution. Our architecture pairs a lightweight, specialized policy model, ContextCurator, with a powerful frozen foundation model, TaskExecutor. Trained via reinforcement learning, ContextCurator actively reduces information entropy in the working memory. It aggressively prunes environmental noise while preserving reasoning anchors, that is, sparse data points that are critical for future deductions. On WebArena, our framework improves the success rate of Gemini-3.0-flash from 36.4% to 41.2% while reducing token consumption by 8.8% (from 47.4K to 43.3K). On DeepSearch, it achieves a 57.1% success rate, compared with 53.9%, while reducing token consumption by a factor of 8. Remarkably, a 7B ContextCurator matches the context management performance of GPT-4o, providing a scalable and computationally efficient paradigm for autonomous long-horizon agents.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。