1 paper · 1 filter
Zeyu Zhang, Bradly C. Stadie
The standard check for contamination in LLM backtests is simple: compare scores before and after the training cutoff. We show this check is uninformative. Four flagship models fail…