text8

Low-hardware language-model research: can useful language-model behaviour be formed on consumer hardware through structure, reuse and search instead of brute force? Small claims, tested honestly.

LLM-Enhanced Text Compression: This project explores using Large Language Models (LLMs) to improve text compression efficiency. It evaluates various LLM architectures (GPT-2, Llama 2, Mistral-7B, Yi-34B) across different context sizes and datasets, implementing a rank-based compression method integrated with conventional algorithms.