Show HN: KV-Cache Grafting – Boosting frozen 12B LLMs to 93.3% AIME accuracy

A new approach called KV-Cache Grafting has been introduced, which aims to enhance the performance of frozen 12B language models to achieve an impressive 93.3% accuracy in AI Model Evaluation (AIME). This method could significantly improve the efficiency and effectiveness of large language models, making them more competitive in various applications. For more details, you can read the full paper here: KV-Cache Grafting.