QLoRA

Fine-Tuning on 4chan Data Boosts Llama 8B and 70B Benchmark Scores

A researcher fine-tuned Llama 8B and 70B on 4chan data and reports both models outperformed their base versions.