Thomas Merth

LazyLLM: DYNAMIC TOKEN PRUNING FOR EFFICIENT LONG CONTEXT LLM INFERENCE

(Dynamic) Prompting might be all you need to repair Compressed LLMs