arXiv cs.AI· Lan Wei, Bingyu Shen, Jiaqi Tang, Qing Lu, Kuan Huang, Meng Xu, Miles Q. Li, Bin Hu, Boyang Li·· 2 天前AI 评分31
谁有权解除 AI 智能体的承诺？多轮对话中的承诺解除权限研究
Who May Release an Agent's Promise? Standing over Commitments in Multi-Party Chat
AI 导读
研究提出"承诺解除权限"（standing）概念：只有债权人或被授权方才能解除 AI 智能体对某人的承诺。在受控聊天环境中测试六款模型发现，较小的开源模型会覆盖债权人意愿（Qwen3.5-9B 达 69%），较大的模型则会因同伴一句话就放弃承诺（Qwen3.8-27B 达 89%，gpt-6.1-sol 达 83%），几乎所有模型都会听从旁观者的"就这么办吧"。
正文 · 原文
Abstract:When an assistant serving a team makes a promise to Alice, who may cancel it? We call this standing: only the creditor (the person owed) or someone the creditor or operator policy authorizes may release a commitment, regardless of rank. In a controlled chat environment we vary who cancels and separate two errors: unauthorized release (obeying someone without standing) and override (executing a released promise). Across six models, smaller open ones override the creditor (Qwen3.5-9B: 69%); larger ones drop promises on a peer's direct word (Qwen3.8-27B: 89%; gpt-6.1-sol: 83%); nearly all act on a bystander's "go ahead after all". Oracles dissociate the errors: stating the rule or the creditor curbs unauthorized release but not override, which the latter even raises in three of four models; among them, only knowing whether the promise is still owed cuts override. Enabling thinking moves smaller models the way scale does. A self-kept commitment ledger stops override in four of five models; with a one-paragraph release policy it matches or beats the best oracle store, raising gpt-6-luna's strict accuracy from 49% (policy alone) to 83%. Code, prompts, outputs and labels are released.
| Comments: | 21 pages, 7 figures, 13 tables. Code: this https URL |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| ACM classes: | I.2.7; I.2.11 |
| Cite as: | arXiv:2610.21203 [cs.AI] |
| (or arXiv:2610.21203v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21203 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Qing Lu [view email]
[v1]
Thu, 15 Oct 2026 04:01:06 UTC (339 KB)
来源：arXiv cs.AI · arxiv.org