ResearchBenchmarksFinding Extraction on RadTimelineFollow75Recall (Finding)GPT-4o61.4864.9968.572.01Mar 24, 2026Evaluation ResultsMethodMethodLinksRecall (Finding)Recall (Reported)Miss RateUnnecessary Finding RateGPT-4oAbsent finding filter=...Absent finding filter=more aggressive2026.0375—615Llama-3.1Absent finding filter=...Absent finding filter=more aggressive2026.0370—332GPT-4oAbsent finding filter=...Absent finding filter=weaker2026.036768525Llama-3.1Absent finding filter=...Absent finding filter=weaker2026.036267340