<?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>제로샷 on TechLLM</title><link>https://techllm.github.io/tags/%EC%A0%9C%EB%A1%9C%EC%83%B7/</link><description>Recent content in 제로샷 on TechLLM</description><generator>Hugo</generator><language>ko</language><lastBuildDate>Thu, 28 May 2026 13:07:00 +0900</lastBuildDate><atom:link href="https://techllm.github.io/tags/%EC%A0%9C%EB%A1%9C%EC%83%B7/index.xml" rel="self" type="application/rss+xml"/><item><title>Uni-LaViRA — 언어·비전·로봇 행동을 하나로 묶는 제로샷 체화 항법</title><link>https://techllm.github.io/posts/uni-lavira-language-vision-robot-navigation-zero-shot/</link><pubDate>Thu, 28 May 2026 13:07:00 +0900</pubDate><guid>https://techllm.github.io/posts/uni-lavira-language-vision-robot-navigation-zero-shot/</guid><description>데이터 없이 네 종류 로봇이 네 종류 탐색 과제를 수행한다. Uni-LaViRA는 MLLM의 자연 출력 공간에 항법 결정 구조를 맞춰 학습 없이 VLN-CE R2R 60.7%, HM3D-v2 77.7%를 달성했다.</description></item></channel></rss>