Cheaply Estimating Inference Efficiency Metrics for Autoregressive Transformer Models

Aug-13-2025, 06:09:20 GMT–Neural Information Processing Systems 

Large language models (LLMs) are highly capable but also computationally expensive.