Wnuan: Staged Post-Training for Question Answering over Proprietary Enterprise Knowledge

  • 类型：arxiv
  • 标识：2608.01862
  • 链接：https://arxiv.org/abs/2608.01862
  • 主分类：engineering
  • 形态：method
  • 被引：0
  • 被引来源：Semantic Scholar
  • S2被引：0
  • 影响力被引：0
  • TLDR：Wnuan, a three-stage pipeline that constructs task-oriented supervision from documents, performs supervised fine-tuning with general-data replay, and applies reinforcement learning to residual errors is presented, describing both the gains and the general-capability cost of staged enterprise adaptation.
  • 待LLM分类：否
  • 标题中文：标题 -> 标题中文：Wnuan：面向专有企业知识问答的分阶段后训练
  • TLDR中文：本文提出 Wnuan 三阶段流程：从文档构建任务导向监督，结合通用数据回放进行监督微调，对残余误差应用强化学习，阐述分阶段企业适配的收益与通用能力代价
  • 来源文件：
  • /inbox/tom/_candidates/2026-08-05-agent-rag-longcontext-candidates.json
  • [S2 enrich]