<?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>Process-Reward-Models on Rajat Patel</title><link>https://rajathpatel23.github.io/tags/process-reward-models/</link><description>Recent content in Process-Reward-Models on Rajat Patel</description><generator>Hugo -- 0.162.1</generator><language>en-us</language><managingEditor>rpatel12@umbc.edu (Rajat Patel)</managingEditor><webMaster>rpatel12@umbc.edu (Rajat Patel)</webMaster><lastBuildDate>Sun, 07 Jun 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://rajathpatel23.github.io/tags/process-reward-models/index.xml" rel="self" type="application/rss+xml"/><item><title>Evidence-Driven Deep Research Agent</title><link>https://rajathpatel23.github.io/posts/deep-research-agent/</link><pubDate>Sun, 07 Jun 2026 00:00:00 +0000</pubDate><author>rpatel12@umbc.edu (Rajat Patel)</author><guid>https://rajathpatel23.github.io/posts/deep-research-agent/</guid><description>How I built visibility into the research process itself — an explicit evidence state, step-level reward signals, and a planner that uses those signals deliberately. Without retraining.</description></item></channel></rss>