ScholarPulse 日报 2026-10-21

2026-10-21 学术简报：2 篇。Agent S4 是一个自主的计算机使用科学家，它通过 meta-agent 从任务代理的轨迹中构建和修订 Evolving Agent Interfaces，利用 population-based evolution 选择泛化良好的接口仓库，在 OSWorld V2 上显著提升了代理性能，达到新的开源 SOTA，性能超过 GPT-5.5 和 Claude Opus，且成本更低。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	Evolving Agent Interfaces	arXiv	2026-10-20	AI-Agent	高
2	Cache Engineering for Agents	arXiv	2026-10-20	RAG	高

重点论文与技术动态

1. Evolving Agent Interfaces

来源：arXiv
日期：2026-10-20
作者/机构：Saaket Agashe, Yuzhe Yang, Gonzalo Gonzalez-Pumariega, Bhuvan Chandra Koduru, Chih-Lun Li, Jiachen Yang
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.AI

一句话结论

Agent S4 是一个自主的计算机使用科学家，它通过 meta-agent 从任务代理的轨迹中构建和修订 Evolving Agent Interfaces，利用 population-based evolution 选择泛化良好的接口仓库，在 OSWorld V2 上显著提升了代理性能，达到新的开源 SOTA，性能超过 GPT-5.5 和 Claude Opus，且成本更低。

核心内容

计算机使用代理受限于人类给定的接口，如逐个输入单元格或打印整个文件来查找功能。
接口和使用知识由人工构建，无法跟上新应用、 workflows 和模型的快速发展。
Agent S4 使接口设计成为代理能力，通过 meta-agent 自主构建和修订接口。

方法与数据

Meta-agent 分析合成训练任务的代理轨迹，构建和修订接口和使用知识的仓库。
使用 population-based evolution 选择表现良好的接口仓库，在 OSWorld V2 上使用 Qwen3.8-27B 模型测试。

价值判断

值得关注：Agent S4 达到 21.3% 二元分数，是固定接口的 3.8 倍，超过 GPT-5.5 (13.0%) 和 Claude Opus 4.8 (20.6%)，成本仅为它们的 4.3% 和 1.5%。
可复用点：开源的代码和演化接口仓库，可用于其他计算机使用代理任务的开发。
局限/待核查：合成任务与真实任务的泛化性；具体模型大小和成本的细节。

摘要

Computer-use agents are limited by the interfaces humans give them. An agent may type a spreadsheet cell by cell when a single line of code would do, or print all two thousand lines of a file through a shell to find one function. Both the interfaces and the knowledge of when to use each are built by hand, which cannot keep pace with new applications, workflows, and models. We argue that interface design should itself be an agent capability. We introduce Agent S4, an autonomous computer-use scientist that builds Evolving Agent Interfaces from task experience. A meta-agent examines task agents' trajectories on synthetic training tasks, without success labels, and constructs and revises interfaces and usage knowledge in a repository. Population-based evolution then selects which repositories to develop further by their performance on validation tasks the meta-agent never sees, favoring changes that generalize beyond the seen tasks. On OSWorld V2, Agent S4, with Qwen3.8-27B and the evolved repository of interfaces, sets a new state of the art in open-source computer-use agents, reaching 21.3% binary score, 3.8 times the same model with fixed interfaces. This is above both GPT-5.5 (13.0%) and Claude Opus 4.8 (20.6%), at about 4.3% and 1.5% of their respective costs. On ScienceBoard, the evolved interfaces improve the success rate from 41.4% to 58.6% with the same model. These results suggest that useful interfaces can be evolved from experience, making autonomous interface design a source of improvement complementary to advances in models and harnesses. We open-source our code at https://github.com/simular-ai/agent-s.

2. Cache Engineering for Agents

来源：arXiv
日期：2026-10-20
作者/机构：Tianyu Fu, Ruoxi Wang, Zihan Min, Zhongqi Ren, Zhihao Li, Tengxuan Liu
主题标签：RAG, arXiv
推荐等级：高
分类：cs.CL

一句话结论

提出缓存工程方案，使LLM智能体无需修改轨迹或重新预填充即可直接控制活动KV缓存，通过drop和repos操作降低成本并提升性能。

核心内容

智能体长轨迹使KV缓存I/O和存储成主导成本，轨迹常超出LLM最大上下文长度。
先前上下文工程修剪或总结历史会修改前缀，导致后续缓存失效并触发昂贵重填充，保留过少上下文还会损害准确性。
缓存工程允许智能体主动drop排除选定缓存、repos调整位置编码，无需修改轨迹。

方法与数据

形式化缓存操作并识别历史依赖性（相同token序列因操作顺序产生不同缓存状态），共同设计AKV引擎：无状态接口兼容现有API，内部管理有状态重用、驱逐与恢复。
工具响应滚动drop策略比阈值触发总结平均基准分数提高1.7分、成本降27.7%；AKV实现最高1.2倍输出吞吐量和1.5倍P95延迟加速；策略递归自改进再降成本7.1%。

价值判断

值得关注：为智能体长轨迹缓存管理提供新范式，显著降本并提吞吐延迟。
可复用点：drop/repos操作、AKV引擎及滚动drop策略，代码已开源。
局限/待核查：摘要未明确具体基准数据集细节。

摘要

Large Language Model (LLM) agents can handle increasingly complex tasks, navigating long trajectories of observations and actions. As these trajectories grow, key-value (KV) cache I/O and storage become the dominant cost. Trajectories can also exceed the LLM's maximum context length, requiring control over active cache growth. Prior work uses context engineering to reduce cache size by trimming or summarizing history. However, modifying the trajectory prefix invalidates subsequent cached states and triggers costly re-prefilling of retained context. Task accuracy can also suffer when too little context is retained to limit re-prefill costs. We propose cache engineering as a new scheme. It allows agents to directly control their active KV cache without modifying the trajectory or re-prefilling retained context. Specifically, agents can proactively use drop to exclude selected cache from future attention and repos to adjust retained keys' position encodings. We formalize these cache operations and identify history dependence as a serving challenge: identical token sequences can yield different cache states under different cache operation sequences, making prefix matching insufficient for correct cache reuse. Thus, we co-design the AKV serving engine. It provides a stateless interface compatible with existing serving APIs, while internally managing stateful cache reuse, eviction, and restoration. Experiments show that a tool-response rolling-drop cache policy improves average benchmark scores by 1.7 points and reduces average cost by 27.7%, compared with conventional threshold-triggered summarization. With AKV, it achieves up to 1.2x the output throughput and a 1.5x speedup in P95 request latency. Recursive self-improvement of cache policies further reduces cost by 7.1% over expert design. Our code is at https://github.com/thu-nics/AKV.