AI Daily AI Daily – 2025-06-24(Evening) AI benchmarkingAI safetyDeep learning model evaluationDynamic AI benchmark XbenchFaithfulSAE feature extractionGemini Robotics On-DeviceLiveCodeBenchLiveCodeBench Pro programming testmultimodal modelsReinforcement learningSlimMoE model compression frameworksparse autoencoderXbench