Shengen Yan

PM-KVQ: Progressive Mixed-precision KV Cache Quantization for Long-CoT LLMs

R2R: Efficiently Navigating Divergent Reasoning Paths with Small-Large Model Token Routing

Distilled Decoding 2: One-step Sampling of Image Auto-regressive Models with Conditional Score Distillation

Mixture of Attention Spans: Optimizing LLM Inference Efficiency with Heterogeneous Sliding-Window Lengths

LV-Eval: A Balanced Long-Context Benchmark with 5 Length Levels Up to 256K

Linear Combination of Saved Checkpoints Makes Consistency and Diffusion Models Better

ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation

MoA: Mixture of Sparse Attention for Automatic Large Language Model Compression

LV-Eval: A Balanced Long-Context Benchmark with 5 Length Levels Up to 256K

Efficient Expert Pruning for Sparse Mixture-of-Experts Language Models: Enhancing Performance and Reducing Inference Costs