Tag
EHR2Trace is a configurable system that converts heterogeneous electronic health records into source-linked, auditable patient events exported to OMOP and MEDS, separating event time from information availability. Across three clinical datasets it converted 846.4M events and detected all 28 injected faults, while showing availability-filtered training improves held-out AUROC (0.829 vs 0.642) and exposing inflated performance (0.965 AUROC) from backdated histories.
This paper audits temporal leakage in financial news NLP benchmarks across multiple models, finding that random splits inflate performance metrics and identifies M&A events as a category with a localized positive signal under chronological evaluation.
This paper shows that the standard pre/post training-cutoff check for temporal leakage in LLM backtesting is uninformative, as recency effects mimic leakage. It proposes new estimators using known cutoffs and matched clean controls to measure leakage and compute adjusted scores, validated on frontier models.