Anyone else share the suspicion that ML rapidly approaching 100% on benchmarks is sometimes due to releasing the test set?
Anyone else share the suspicion that ML rapidly approaching 100% on benchmarks is sometimes due to releasing the test set?