arXiv cs.MA· Yaxiao Liu (PwC China AI Center), Pengbo Liu (PwC China AI Center), Yiwen Liu (PwC China AI Center), Yihua Guan (PwC China AI Center), Jiaxing Song (Tsinghua University), Xingjian Li (PwC China AI Center)·· 2 天前AI 评分22
从迁移到校准：让 AI 智能体对齐部署需求
From Migration to Calibration: Aligning Agents with Deployment Requirements
AI 导读
论文提出面向部署目标的智能体校准框架，主张校准应让智能体满足目标环境需求，而非模仿来源或追求通用最优。框架定义三类标准族（基础能力与可操作性、技术与专业精度、用户上下文），并用版本化映射区分需保留、转换、新增或淘汰的需求。作者强调这是方法论提案，仅含部分调和的试点证据，确认性验证与超 80% 零样本任务上下文接受目标仍未完成。
正文 · 原文
Abstract:Deploying, migrating, or scaling an agent changes its model, support system, application, and users. Calibration should align the agent with destination requirements without requiring source imitation or universal superiority. We propose three standard families: basic capability and operational reachability; technical and professional precision, including terminology, units, reporting periods, program fidelity, and report dependencies; and user context covering roles, use cases, authorized values, permissions, and budgets. A versioned mapping distinguishes requirements to preserve, transform, add, or retire. Four instruments---benchmarks, positive/negative examples, skills, and role-conditioned cost/utility---supply tunable reward channels within a fixed acceptance contract. Consultants annotate methods, branches, bounded loops, checkpoints, relative comparisons, and report refinement; a registry records applicable evidence. Adaptation revises configurations or trains a separate controller or skill while freezing the foundation executor. Independent hard gates govern qualification; RL and preference methods optimize candidates without supplying factual truth. Qualification and held-out improvement over the unchanged target are separate outcomes. Supplied intent labels permit a limited paired recount: small prompt/retry gains coexist with regressions and failed gates. FinQA supplements match prior summaries and retain provenance conflicts. These observations motivate destination-specific evaluation but do not establish calibration efficacy. We specify matched-budget tests, clean reload, complete-denominator reporting, and dependency revalidation. This is a methodological proposal with partially reconciled pilot evidence; confirmatory validation and the greater-than-80% zero-shot task-context acceptance target remain outstanding.
| Comments: | 64 pages, 10 figures, 11 tables. Revised destination-oriented calibration framework and a partially reconciled pilot evidence audit. Methodological proposal; no confirmatory calibration or RL efficacy results. Matched-budget validation and the automatic calibration tool remain future work |
| Subjects: | Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA); Software Engineering (cs.SE) |
| Cite as: | arXiv:2609.35149 [cs.AI] |
| (or arXiv:2609.35149v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.35149 arXiv-issued DOI via DataCite |
Submission history
From: Yaxiao Liu [view email]
[v1]
Mon, 28 Sep 2026 13:31:46 UTC (807 KB)
[v2]
Fri, 2 Oct 2026 13:41:01 UTC (960 KB)
[v3]
Thu, 15 Oct 2026 00:48:35 UTC (1,184 KB)
来源：arXiv cs.MA · arxiv.org