Quantifying LLM Base Model Quality through Compression

2025‑04‑29

Introduces a Danish-centric benchmark and leaderboard that grades base LLMs by compression efficiency on fresh Danish, English and Python text.