Automating the Design of Embodied Agent Architectures
- 类型：arxiv
- 标识：2606.30111
- 链接：https://arxiv.org/abs/2606.30111
- 主分类：agent
- 形态：position
- 被引：1
- 被引来源：Semantic Scholar
- S2被引：1
- OpenAlex被引：0
- 影响力被引：0
- TLDR：This work evaluates three AAS variants across four embodied executors spanning vision-language navigation, embodied question answering, and language-conditioned manipulation, and shows that architecture-level search can produce deployable and directional success-rate gains on embodied tasks, while one apparent high-scoring candidate is rejected as leak-bearing.
- OpenAlex ID：W7166698998
- OpenAlex DOI：10.48550/arxiv.2606.30111
- DOI：10.48550/arxiv.2606.30111
- DOI来源：OpenAlex
- 开放获取：green
- 开放获取链接：https://doi.org/10.48550/arxiv.2606.30111
- OpenAlex更新：2026-07-19
- 待LLM分类：否
- 标题中文：具身智能体架构设计的自动化
- TLDR中文：本文在视觉语言导航、具身问答和语言条件操控任务上评估了三种 AAS 变体，覆盖四个具身执行器，结果表明架构级搜索能在具身任务上产生可部署且具有方向性的成功率提升，而其中一个看似得分较高的候选因存在泄漏而被判定为无效。
- 来源文件：
- /inbox/tom/_candidates/2026-07-09-agent-rag-longcontext-candidates.json
- [S2 enrich]
- [OpenAlex backfill]