Muon Optimizer Significantly Accelerates Grokking in Transformers: Microsoft Researchers Explore Optimizer Influence on Delayed Generalization Revisiting the Grokking Challenge In recent years, the phenomenon of grokking —where models exhibit a delayed yet sudden transition from memorization to generalization—has prompted renewed investigation into training dynamics. Initially observed in small algorithmic tasks like modular arithmetic, grokking reveals… →

LLMs Can Now Learn without Labels: Researchers from Tsinghua University and Shanghai AI Lab Introduce Test-Time Reinforcement Learning (TTRL) to Enable Self-Evolving Language Models Using Unlabeled Data Despite significant advances in reasoning capabilities through reinforcement learning (RL), most large language models (LLMs) remain fundamentally dependent on supervised data pipelines. RL frameworks such as RLHF have… →

Open-Source TTS Reaches New Heights: Nari Labs Releases Dia, a 1.6B Parameter Model for Real-Time Voice Cloning and Expressive Speech Synthesis on Consumer Device The development of text-to-speech (TTS) systems has seen significant advancements in recent years, particularly with the rise of large-scale neural models. Yet, most high-fidelity systems remain locked behind proprietary APIs and… →