Community discussion: improving language model inference speed

A Reddit user asked for recent papers on LLM inference architectures and proposed a Bell Labs-style 'lunch table' discussion on speeding up inference. The poster, who says they have worked with LLMs for two months, shared no specific technical findings.