Karsten Mueller

Activation- and Influence-Aware Ranks (AIR): Function-Preserving SVD Compression for LLMs

An SVD-based LLM compression framework that guides each weight matrix's low-rank approximation with a backward-signal influence metric, which improves perplexity over SVD-LLM(W) by>18% at<=60% parameter retention, matches its quality with ~90% less calibration data, and turns parameter savings into FLOP, peak-memory, a...