金融监管领域的 Harness 实践：让知识与数据驱动 Agent 稳定运行
摘要
中软融鑫总工程师于浩军 在 AICon 2026 深圳站分享了金融监管报送场景下的 Harness 实践，用知识底座、数据底座、受控状态机与全链路可追溯，把容易发散的大模型 Agent 改造成“宁可不答、不能答错”的严格受控系统。该场景对准确率要求接近 100%，远高于一般业务里 97% 就算不错的水平。适合在强监管、高合规行业落地 Agent 的团队参考。
近 7 天热度
相关推荐
[分享创造] 关于自己的 agent-harness 的实践与迭代过程中一些好玩的点
affaan-m/ECC
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
从 Harness 到 Loop：Agent 如何走向可靠交付与持续进化？
affaan-m/ECC
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
ruvnet/ruflo
🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated