arXiv cs.CL· Petmongkon Pornpichitsuwan, Krittin Pachtrachai, Touchapon Kraisingkorn·· 3 天前AI 评分30
训练语音智能体遵循未见过的债务催收监管流程
Following the Flow the Law Requires: Training Voice Agents to Transfer to Unseen Regulated Procedures in Debt Collection
AI 导读
研究者提出 FlowSpec，用一份机器可读规范为每次部署生成策略指令、门控工具、提供监督并定义测试，智能体只能说经批准的目录语句。基于 297 个随机流程的 2,506 段对话微调 9B 参数模型，在四个未见部署流程上，流程固定的文档化检查通过率从 67–71% 升至 93–100%，场景通过率从 15/43 升至 40/43。对共享规范的审计发现 11 处缺陷，其中 3 处未被其自身指标暴露。
正文 · 原文
Abstract:Voice agents built on large language models (LLMs) now handle customer calls, but in regulated applications completing the task is not enough: the agent must follow a conversational flow set by law. In debt collection, it must confirm it is speaking to the debtor before mentioning the debt, identify itself and the creditor, state the amount, and avoid prohibited language. Every creditor turns these rules into its own procedure and revises it. We ask whether a small, self-hosted policy can be trained once and then follow regulated procedures it has never seen. One machine-readable specification per deployment, FlowSpec, generates the policy instruction, gates the tools, supplies the supervision and defines the tests; the agent speaks only approved catalogue sentences. We fine-tune a 9-billion-parameter model on 2,506 conversations from 297 randomized procedures and evaluate a Thai outbound-call agent on four unseen deployment procedures (three debt collection, one non-debt), one from a wholly withheld lineage. Documented checks, fixed by the procedure, rise from 67--71 percents to 93--100 percents, while grounded checks, fixed by the conversation, already pass at 97--100 percents before training; scenario pass rate rises from 15/43 to 40/43. An audit of the shared specification finds 11 defects, 3 of which its own metric did not expose.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.18881 [cs.CL] |
| (or arXiv:2610.18881v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.18881 arXiv-issued DOI via DataCite |
Submission history
From: Krittin Pachtrachai PhD [view email]
[v1]
Thu, 1 Oct 2026 03:08:52 UTC (74 KB)
来源：arXiv cs.CL · arxiv.org