4 ms·LLM Inference Throughput Rises 4.5x with Parallel Verification2 points by sebastianperezr 5mo ago