1 paper · 1 filter
Bryan Kelly, Semyon Malamud, Johannes Schwab +1
Large language models trained on unrestricted internet corpora inevitably embed information from the future, introducing lookahead bias that compromises the validity of backtests a…