<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Financial on Yuan's Blog</title><link>https://liyuan.org/tags/financial/</link><description>Recent content in Financial on Yuan's Blog</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sun, 10 May 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://liyuan.org/tags/financial/index.xml" rel="self" type="application/rss+xml"/><item><title>Financial RAG Agent Optimization:Methods, Cases, and Data</title><link>https://liyuan.org/posts/ai/rag-improvement-via-18-failures/</link><pubDate>Sun, 10 May 2026 00:00:00 +0000</pubDate><guid>https://liyuan.org/posts/ai/rag-improvement-via-18-failures/</guid><description>This project details the refinement of an agentic RAG system for financial Q&amp;amp;A, boosting test accuracy from &lt;strong>0.871 to ~0.919&lt;/strong> by systematically diagnosing 18 failure cases. Rather than blind model tuning, the author prioritized a &amp;quot;diagnose-first&amp;quot; approach: resolving &amp;quot;judge-side&amp;quot; discrepancies with deterministic numeric prefiltering, then implementing structural improvements like query translation, anti-refusal checks, and a five-layer fix for superlative ambiguities. The results highlight that while prompt-based reflection is helpful, structural, schema-enforced changes offer superior reliability. Ultimately, the author demonstrates engineering pragmatism by consciously leaving eight failures unfixed—due to dataset noise or unfavorable ROI—distinguishing between &amp;quot;fixing everything&amp;quot; and strategic, production-oriented optimization.</description></item></channel></rss>