今日速览
| 序号 | 标题 | 来源 | 日期 | 主题 | 推荐等级 |
|---|---|---|---|---|---|
| 1 | The Athlete as a Scientist of One: Agentic Bayesian Experimental Design for Individualized Physiology from Consumer Wearables | arXiv | 2026-10-29 | AI-Agent | 高 |
| 2 | Don’t Lose the Thread: Empowering Long-Horizon LLM Agents with Cognitive Resource Self-Allocation | arXiv | 2026-10-29 | RAG | 高 |
重点论文与技术动态
1. The Athlete as a Scientist of One: Agentic Bayesian Experimental Design for Individualized Physiology from Consumer Wearables
- 来源：arXiv
- 日期：2026-10-29
- 作者/机构：Behnam Moradi
- 主题标签：
AI-Agent,arXiv - 推荐等级：高
- 分类：cs.RO, cs.AI
一句话结论
EpiCoach利用机理数字孪生与证据门控注册表，将可穿戴设备数据转化为可审计、可证伪的个体化贝叶斯训练实验设计。
核心内容
- 将运动员作为N-of-1研究对象，所有假说必须在训练前输出概率预测，且仅依据统计拟合证据更新信念。
- 整合心率滞后响应、体能-疲劳及下坡技术模型构建数字孪生，在负荷与安全约束下以最大化信息增益规划训练。
- 实测中记录了20项预测并实现校准，所有修正均被留存，且安全员曾介入并2次否决智能体决策。
方法与数据
- 基于一名越野山地车手18个月的心率、GPS及气压海拔数据，结合滚动原点回放与实时前瞻周期进行验证。
价值判断
- 值得关注：构建了首个具备可审计与可证伪特性的个体化健康智能体闭环实验框架。
- 可复用点：开源的假说注册表、评测基准及实验协议。
- 局限/待核查：路段耗时预测未超越标准基线，智能体闭环表现目前仅持平静态模型。
摘要
Large language model (LLM) agents for personal health read wearable data and recommend, but do not learn about the individual in a way that can be audited or falsified. We present EpiCoach, an agentic system that treats one athlete, a cross-country mountain biker, as the subject of accountable N-of-1 science. Claims about the athlete live in an evidence-gated hypothesis registry: each must issue a probabilistic prediction about a future session before it occurs, and its belief moves only through computed evidence from a logged statistical fit. Predictions come from a mechanistic digital twin inferred from heart rate, GPS and barometric altitude: a lagged heart-rate response model, a fitness-fatigue model with learned time constants, and a latent technical-descending skill state, measured on discovered trail segments. The agent prescribes the next training week as a constrained experiment maximizing information gain subject to load and safety constraints, around two instrument rides. We evaluate the system two ways over eighteen months of data. A rolling-origin replay benchmark shows the twin beats standard forecasters on heart rate but not segment time, and the agentic loop matches a static model. A live prospective phase runs the weekly cycle: in its first three weeks the ledger sealed 20 predictions, scored 14, the descent forecast calibrated after one correction, the climb forecast pessimistic as the athlete improved, and the safety officer vetoed the agent twice. Every correction is preserved in the registry, a template for auditable, falsifiable personalization in digital health. We release the registry, benchmark and protocol.2. Don’t Lose the Thread: Empowering Long-Horizon LLM Agents with Cognitive Resource Self-Allocation
- 来源：arXiv
- 日期：2026-10-29
- 作者/机构：Zheng Zhu, Xiaozhuan Liang, Xi Chen, Bei Yu, Jiaya Jia
- 主题标签：
RAG,arXiv - 推荐等级：高
- 分类：cs.CL
一句话结论
论文提出认知资源自分配框架 CORAL，通过让智能体自主保存关键信息检查点并重置杂乱上下文，有效解决了长程任务中的认知过载，同时大幅降低了推理成本。
核心内容
- 针对长程任务中因上下文膨胀造成注意力稀释的问题，支持智能体主动优化工作记忆。
- 智能体通过工具自适应保存进度与已验证事实，清空冗余上下文并从干净状态继续推理。
- 在 GAIA、多跳问答及复杂的 BrowseComp 基准上均展现出强泛化性，并显著优于现有基准。
方法与数据
- 方法：设计了智能体可调用的工作记忆管理工具集，并使用多回合智能体强化策略优化算法（Multi-episode Agentic Reinforced Policy Optimization）增强检查点能力。
- 数据：在 GAIA（8B 模型达 42.8% 准确率，Level 2/3 提升显著）、多跳知识问答及 BrowseComp（使某专有商业智能体提升 8.0 分）上评测，输入 token 较 ReAct 减少 6-7 倍。
价值判断
- 值得关注：在长程高难任务上兼顾性能提升与算力节约，输入 token 消耗降低至 ReAct 的 1/6 到 1/7。
- 可复用点：通过保存检查点配合上下文自适应重置的工作记忆管理工具机制。
- 局限/待核查：摘要未明确具体的多跳问答基准名称，以及所测试的专有商业智能体具体型号。