Your question is LLM Benchmark for Gradient Explosion. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How would you design a benchmark to evaluate an LLM's ability to identify and fix a gradient explosion issue in a deep neural network?