MMODELYST
People/Junlong Ke
PPL

Junlong Ke

Papers (3)
FlexDraft: Flexible Speculative Decoding via Attention Tuning and Bonus-Guided CalibrationRespecting Self-Uncertainty in On-Policy Self-Distillation for Efficient LLM ReasoningEvoStreaming: Your Offline Video Model Is a Natively Streaming Assistant