AI Daily – 2025-06-01(Evening)
AI-generated CUDA kernelsFLUX.1 Kontext image generationGTA and GLA attention mechanismsHuawei's Pangu Ultra MoE model efficient training systemMaskSearch pre-training frameworkNankai University and UIUC optimize AI search agent efficiencyPangu Ultra MoE modelRISEBench evaluation benchmarkSearchAgent-X frameworkShanghai AI Lab's RISEBench multimodal evaluationStanford University AI-generated CUDA kernels outperform humansTON selective reasoning frameworkTri Dao, creator of Mamba, proposes GTA and GLA attention mechanisms