06RESEARCH PAPERPreprint
Hot-Brain Observability for Deployed AI Agents
發表日期
摘要
ABSTRACTZenodo 原文SPVE's hot-brain layer is a monitoring instrument, not a biological claim. It records functional state variables such as attention load, working-memory pressure, predictive error, executive control, reward drive, social model update, self-model continuity, and pragmatic intent precision. The aim is practical: when many agents are working at once, an operator needs to see overload, pressure, repair, and readiness before the next outcome arrives.
全文 PDF
FULL TEXT1 個檔- 下載 PDFpaper.pdf・43 KB・本站
PDF 與 Zenodo 上的檔案相同（大小與 md5 已核對）。
論文資料
DETAILSZenodo 紀錄 20743904- 作者
- Chen, Ho Yiing
- 發表日期
- 資源類型
- Preprint
- DOI
- 10.5281/zenodo.20743904
- Zenodo
- zenodo.org/records/20743904
- 授權
- Creative Commons Attribution 4.0 International
作者 Chen, Ho Yiing 是織田紀香的本名（陳禾穎）英文拼寫。標題、摘要、關鍵字與授權照 Zenodo 原文。
關鍵字
KEYWORDS24 個- SPVE
- AI agents
- persistent agents
- longitudinal agents
- multi-agent systems
- agent ecology
- agent observability
- lobster agents
- behavioral telemetry
- agent monitoring
- workflow ecology
- agent evaluation
- hot-brain parameters
- neuro-inspired observability
- functional monitoring
- attention load
- working memory pressure
- predictive error
- executive control
- reward drive
- social model update
- self-model continuity
- survival pressure
- live dashboard