SALT: Salience-Aware Lexical Trie for Long-Context Compression
Model-agnostic extractive prompt compression framework using sentence-frequency-ordered lexical trie to prevent theme collapse under tight budgets
1 post tagged with "inference-efficiency"