{"ok":true,"trend":{"id":900402,"platform":"mastodon","region":"global","key":"meta ai、ループ型moeの設計則を提案 推論課題で約2倍規模のモデルに匹敵 | 財経新聞 https://www. yayafa.com/2901872/ # agenticai # ai # artificialgeneralint","title":"Meta AI、ループ型MoEの設計則を提案 推論課題で約2倍規模のモデルに匹敵 | 財経新聞 https://www. yayafa.com/2901872/ # AgenticAi # AI # ArtificialGeneralInt","url":"https://jforo.com/@yayafa/117378818370050247","first_seen":"2026-10-03T20:24:23.583461Z","last_seen":"2026-10-03T21:47:16.348997Z","last_rank":11,"peak_rank":8,"last_volume":2,"peak_volume":2,"seen_count":2,"score":0.814375,"category_hint":"ai","section":"technology","category":"ai","summary":"Meta AI researchers have proposed design principles for a looped mixture-of-experts (MoE) architecture, reporting that it performs on reasoning tasks on par with models roughly twice its size. The approach reuses expert layers in a loop rather than scaling parameters, offering a path to more efficient large language models. The news, reported by Japanese outlet Zaikei Shimbun, is being shared among AI researchers and enthusiasts tracking efficiency gains in model design.","why":"Interest in techniques that let smaller AI models match much larger ones on reasoning benchmarks, a key cost and capability question in AI right now.","tone":"neutral","entities":["Meta AI","mixture-of-experts","Llama"],"summarized_at":"2026-10-03T20:26:07.882585Z","meta":{"tag":"artificialintelligence","via":"scan","kind":"status","lang":"ja","instance":"mastodon.social","tag_uses":535},"nw":null,"promo":null,"kind":null,"importance":null,"hidden":false,"hide_reason":null,"judged_at":null,"title_en":"Meta AI proposes looped MoE design matching models twice its size","section_name":"Technology","category_name":"AI","timeline":[{"captured_at":"2026-10-03T20:24:23.583461Z","rank":8,"volume":2},{"captured_at":"2026-10-03T21:47:16.348997Z","rank":11,"volume":2}],"posts":[{"platform":"mastodon","url":"https://jforo.com/@yayafa/117378818370050247","author":"yayafa@jforo.com","title":null,"snippet":"Meta AI、ループ型MoEの設計則を提案 推論課題で約2倍規模のモデルに匹敵 | 財経新聞 https://www. yayafa.com/2901872/ # AgenticAi # AI # ArtificialGeneralIntelligence # ArtificialIntelligence # IT・サイエンス # LLAMA # Meta # MetaAI # エージェント型AI # ロボット・人工知能 （AI） # 人工知能 # 汎用人工知能","posted_at":"2026-10-03T20:17:09Z","likes":2}],"elsewhere":[],"window":"7d"}}