<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Embeddings on skulldc</title><link>https://skulldc.net/tags/embeddings/</link><description>Recent content in Embeddings on skulldc</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Wed, 04 Nov 2026 20:26:21 +0000</lastBuildDate><atom:link href="https://skulldc.net/tags/embeddings/index.xml" rel="self" type="application/rss+xml"/><item><title>FactEmbed factorizes MLLM embeddings for composed image retrieval</title><link>https://skulldc.net/2026/11/factembed-factorizes-mllm-embeddings-for-composed-image-retrieval/</link><pubDate>Wed, 04 Nov 2026 20:26:21 +0000</pubDate><guid>https://skulldc.net/2026/11/factembed-factorizes-mllm-embeddings-for-composed-image-retrieval/</guid><description>A KAIST AI paper makes an MLLM imagine the requested image edit through factorized latents before building retrieval embeddings, and reuses those latents for cheap reranking.</description></item></channel></rss>