<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Stable-RAG on Antonio V. Franco</title><link>https://antoniovfranco.com/tags/stable-rag/</link><description>Recent content in Stable-RAG on Antonio V. Franco</description><generator>Hugo</generator><language>en</language><lastBuildDate>Thu, 07 May 2026 12:00:00 -0300</lastBuildDate><atom:link href="https://antoniovfranco.com/tags/stable-rag/index.xml" rel="self" type="application/rss+xml"/><item><title>What the NOA-Star Simulation Reveals About Transient Classification</title><link>https://antoniovfranco.com/posts/2026/noa-star-stable-rag-transient-classification/</link><pubDate>Thu, 07 May 2026 12:00:00 -0300</pubDate><guid>https://antoniovfranco.com/posts/2026/noa-star-stable-rag-transient-classification/</guid><description>The NOA-Star simulation of Stable-RAG on astronomical transient classification achieves 98.9% accuracy across 90 events, but the single error and the low-stability events reveal more than the successes. Stability scores expose where reasoning is fragile, even when the answer is correct.</description></item><item><title>Stable-RAG Benchmark: Your RAG System Lies Depending on Document Order.</title><link>https://antoniovfranco.com/posts/2026/stable-rag-benchmark/</link><pubDate>Mon, 27 Apr 2026 12:00:00 -0300</pubDate><guid>https://antoniovfranco.com/posts/2026/stable-rag-benchmark/</guid><description>Benchmark comparing conventional RAG vs Stable-RAG reveals that reordering retrieved documents causes hallucination. Stable-RAG reduces Permutation Sensitivity Rate by 37.8% using spectral clustering and DPO alignment on Qwen3-4B.</description></item></channel></rss>