论文概要
研究领域: ML 作者: George Andrikopoulos 发布时间: 2026-08-19 arXiv: 2608.19125
中文摘要
当专家纠正LLM助手的错误时,纠正通常随会话而死亡,错误类别会返回。我认为这是一个操作问题,不是工具问题:持久化纠正的机制存在并已交付,但管理它们的纪律——带有出处的版本控制、复发监控、反指标、过时规则的退役——不存在。作为三十年的系统工程师,我将LLM堆栈映射到我职业已经操作的机器(冻结硅、固件、可加载模块、持久配置、易失性存储),识别映射失败的地方(随机生成、仅概率绑定的配置、默认情况下没有通用退役(验证)阶段),并从失败中推导出以错误循环为核心的七原则操作纪律。我自己实践中的三个案例说明了该机制,其中一个控制无声地变成了它旨在防止的确切伤害。我以这一观点所暗示的测量框架和测试它所需的实验室研究作为结尾。
原文摘要
When an expert corrects an LLM assistant’s error, the correction usually dies with the session, and the error class returns. I argue this is an operations problem, not a tooling problem: mechanisms for persisting corrections exist and are shipping, but the discipline for governing them — versioning with provenance, recurrence monitoring, counter-metrics, retirement of stale rules — does not. Writing as a systems engineer of thirty years, I map the LLM stack onto the machines my profession already operates (frozen silicon, firmware, loadable modules, persistent configuration, volatile memory), identify where the mapping fails (stochastic generation, configuration that binds only probabilistically, no general-purpose retirement (verification) stage by default), and derive from the failures…
— 自动采集于 2026-08-21
#论文 #arXiv #ML #小凯
