Embarrassingly simple self-distillation improves code generation 4 months ago
Think of it less like a test suite and more like an exam. If you're trying to differentiate between the performance of different people/systems/models, you need to calibrate the difficulty accordingly.
When designing a benchmark, a pass rate of roughly 50% is useful because it gives you the most information about the relative performance of different models. If the pass rate is 90%+ too often, that means the test is too easy: you're wasting questions asking the model to do things we already know it can do, and getting no extra information. And if it's too low then you're wasting questions at the other end, trying to make it do impossible tasks.