06RESEARCH PAPERPreprint
Dialogue as a Behavioral Sensor for Agent Cognition
發表日期
摘要
ABSTRACTZenodo 原文In SPVE, dialogue is not only output. It can become an early sensor for agent state. This draft defines a language-state join: message length, question rate, repair phrases, contradiction markers, direct address, self-reference, and action verbs are aligned with hot-brain and outcome windows. The useful claim is not that language is mind. It is that language may move before settlements reveal success or failure.
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 20743916- 作者
- Chen, Ho Yiing
- 發表日期
- 資源類型
- Preprint
- DOI
- 10.5281/zenodo.20743916
- Zenodo
- zenodo.org/records/20743916
- 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
- dialogue sensor
- language features
- behavioral sensor
- repair phrases
- question rate
- direct address
- self-reference
- action verbs
- state-language alignment
- lead-lag analysis
- conversation telemetry
- agent cognition