TY  - RPRT
TI  - Lexical Prompt Compression for Large Language Models: A Training-Free, Deterministic Pipeline with Empirical Pareto Analysis Across Eleven Task Categories
AU  - Shamin Chokshi
PY  - 2026
UR  - https://arxiv.org/abs/2609.13154
ID  - 2609.13154
ER  -