Comment by adityamishra241

How do you handle task distribution and prevent the benchmark from favoring models that are tuned specifically to these 534 questions?