{"type":"post","id":"trained-via-supervised-cold-start-plus-adagrpo-reinforcement-5f960f","slug":"trained-via-supervised-cold-start-plus-adagrpo-reinforcement-5f960f","title":"Trained via supervised cold start plus AdaGRPO reinforcement learning with mode-relative advantage.","description":"A team from Fudan University, Alibaba, and the University of Southern California proposes AdaVaR, a training framework that teaches a vision-language model to use both text-based and visually grounded reasoning, then select between them by context.","retrieval_nugget":"A team from Fudan University, Alibaba, and the University of Southern California proposes AdaVaR, a training framework that teaches a vision-language model to use both text-based and visually grounded reasoning, then select between them by context. The paper reports strong benchmark results, but they remain author-reported research claims rather than an independently verified product capability.","published_at":"2026-11-11","updated_at":"2026-11-11","record_date":"2026-11-11","date_kind":"published_at","topics":["models","multimodal"],"entities":["AdaVaR","AdaGRPO","Mixture-of-Visual-Thoughts","Qwen2.5-VL","Fudan University","Alibaba Group Holding Limited","University of Southern California"],"related_patterns":["model-answers-become-task-interfaces"],"novelty_label":"new_fact","what_changed":"A new visual-reasoning framework trains AdaVaR with supervised cold-start followed by AdaGRPO reinforcement learning to select reasoning modes based on context.","prior_refs":["https://newruntime.com/posts/learning-new-skills-without-forgetting-self-distillation-vs-4031dd/"],"source_url":"https://arxiv.org/html/2509.22746","source_urls":["https://arxiv.org/html/2509.22746"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:trained-via-supervised-cold-start-plus-adagrpo-reinforcement-5f960f","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/trained-via-supervised-cold-start-plus-adagrpo-reinforcement-5f960f/","markdown":"https://newruntime.com/posts/trained-via-supervised-cold-start-plus-adagrpo-reinforcement-5f960f.md","json":"https://newruntime.com/posts/trained-via-supervised-cold-start-plus-adagrpo-reinforcement-5f960f.json"},"markdown_url":"https://newruntime.com/posts/trained-via-supervised-cold-start-plus-adagrpo-reinforcement-5f960f.md","next_reads":[{"type":"topic","path":"/topics/models/","url":"https://newruntime.com/topics/models/","title":"Models","reason":"Explore the models topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/models.md","json_url":"https://newruntime.com/topics/models.json"},{"type":"topic","path":"/topics/multimodal/","url":"https://newruntime.com/topics/multimodal/","title":"Multimodal","reason":"Explore the multimodal topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/multimodal.md","json_url":"https://newruntime.com/topics/multimodal.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/sources/","url":"https://newruntime.com/sources/","title":"Sources","reason":"See the original sources this archive cites.","media_type":"text/html"}]}