Yandex Research · 5 records

dateidtitleauthorsvenue↑
2026-11-032611.03344FLAMIA: FLAsh MIcroscaling Attention for Fast and Causally Safe Training on BlackwellFedor Velikonivtsev et al.arXiv18
2026-10-062610.06648Representation-Space MMD for Diffusion Language ModelsIlya Drobyshevskiy et al.arXiv26
2026-09-302609.35427LLMs are General Asynchronous AgentsGeorge Yakushev et al.arXiv63
2026-07-162605.16147Registers Matter for Pixel-Space Diffusion TransformersNikita Starodubcev et al.arXiv18
2026-06-302606.30634One-Step Gradient Delay is Not a Barrier for Large-Scale Asynchronous Pipeline Parallel LLM PretrainingPhilip Zmushko et al.arXiv22