Paper Review

A reading vault for efficient AI

Daily reviews of papers on model architecture, post-training, model compression & lightweighting, and on-device AI — currently centered on KV cache quantization and hybrid (full + linear) attention — auto-collected, human-curated, and linked like a vault, one paper a day. model architecture, post-training, model compression·경량화, on-device AI 논문을 매일 한 편씩 리뷰한다 — 지금은 KV cache quantization과 hybrid (full + linear) attention이 중심 — 자동 수집하고 직접 큐레이션해 vault처럼 서로 연결해 둔다.

drag · zoom · hover · click to open

Reading list

Papers by topic

KV Cache & Compression 20

Hybrid & Linear Attention 8

Diffusion LLMs 13

On-device AI 5

Architecture 9

Compression 13

Serving 17