Where we got our idea from
Speculative Decoding with Complementary Quantization Schemes
Accelerating Language Generation through Diffusion
Less is More for Reasoning
Large Language Diffusion Models