Richard Socher - Measuring Language Model Performance with Perplexity (via searchlore.ai)