English summariesThese are short answers, not translations. The Chinese pages hold the derivations, worked examples, and citations.Why is FlashAttention faster?How does PagedAttention save KV memory?What is RadixAttention?Why is EAGLE faster than Medusa?AWQ or GPTQ?FP8 or FP4?MQA, GQA, or MLA?Why did a vLLM GPU stay at 100%?vLLM, SGLang, or TensorRT-LLM?Where do the quantization notes start?