<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Robot Safety | Harshitha B R</title><link>https://harshithabr.github.io/tags/robot-safety/</link><atom:link href="https://harshithabr.github.io/tags/robot-safety/index.xml" rel="self" type="application/rss+xml"/><description>Robot Safety</description><generator>Hugo Blox Builder (https://hugoblox.com)</generator><language>en-us</language><lastBuildDate>Thu, 01 Jan 2026 00:00:00 +0000</lastBuildDate><image><url>https://harshithabr.github.io/media/icon_hu68170e94a17a2a43d6dcb45cf0e8e589_3079_512x512_fill_lanczos_center_3.png</url><title>Robot Safety</title><link>https://harshithabr.github.io/tags/robot-safety/</link></image><item><title>SAFECAST — Robust Failure Detection for VLA Policies</title><link>https://harshithabr.github.io/project/safecast/</link><pubDate>Thu, 01 Jan 2026 00:00:00 +0000</pubDate><guid>https://harshithabr.github.io/project/safecast/</guid><description>&lt;p>Developed a runtime failure-detection framework for Vision-Language-Action robot policies using perturbation-augmented hidden-state probes and conformal calibration. Evaluated &lt;strong>π₀, π₀-FAST, and OpenVLA&lt;/strong> under visual, language, and initial-state distribution shifts across simulation and real-world manipulation.&lt;/p></description></item><item><title>SAFECAST: Robust Failure Detection for VLA Policies with Contrast-Set Training and Calibration</title><link>https://harshithabr.github.io/publication/safecast/</link><pubDate>Thu, 01 Jan 2026 00:00:00 +0000</pubDate><guid>https://harshithabr.github.io/publication/safecast/</guid><description/></item></channel></rss>