Efficient Prompt Compression with Evaluator Heads for Long-Context Transformer Inference

Re-assign community