<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Kartik Goyal</title><description>Research, writing, and projects by Kartik Goyal.</description><link>https://kartikgoyal.ai/</link><item><title>On-Policy Self-Distillation: Continual Learning from Production Feedback</title><link>https://kartikgoyal.ai/research/on-policy-self-distillation/</link><guid isPermaLink="true">https://kartikgoyal.ai/research/on-policy-self-distillation/</guid><description>Can an agent learn from the corrections, clarifications, and tool errors it already receives, and how close does that get to expensive, hand-built training pipelines?</description><pubDate>Thu, 06 Aug 2026 00:00:00 GMT</pubDate></item></channel></rss>