PPL
Xueqi Cheng
Papers (6)
The Stability of Singular Distribution: A Spectral Perspective on the Two-Phase Dynamics of Language Model Pre-trainingExtra-Merge: Tracing the Rank-1 Subspace of Model Merging in Language Model Pre-TrainingSOMA: Efficient Multi-turn LLM Serving via Small Language ModelLatentRouter: Can We Choose the Right Multimodal Model Before Seeing Its Answer?ReAD: Reinforcement-Guided Capability Distillation for Large Language ModelsWhen and Why Grouping Attention Heads Accelerates Muon Optimization