Tell me about a time a research project or model architecture failed. How did you diagnose the failure, and what did you learn from it?
