Guide-yakabatana mibvunzo · Pakati Nhanho

Benchmark Contamination Quiz

Tests how benchmark items leak into training data, detection methods like n-gram overlap and Min-K% Prob, and mitigations such as canaries and time splits.

Nzira dzekutungamira dzinoenderanaBenchmark Contamination
Mubvunzo 1 ye 8

Chii chinonzi benchmark contamination?