ChronoTQA - Phenopackets Onset (No Retrieval): leaderboard
Questions asking at what age a phenotype typically presents in a given disease, grounded in GA4GH Phenopackets case data: the 147 Phenopackets onset questions of ChronoTQA, released with the ChronoMedKG knowledge graph (Ahmed et al., 2026). Models answer from memory without retrieval, and answers are scored as accuracy in percent with a calibrated age-range overlap rubric (tolerance of half the gold range, at most two years), higher is better. The authors note that Phenopackets has been public since 2019, so these scores are bounded by what the models may have seen in pretraining.
Metric: Accuracy (%; 147 Phenopackets-grounded onset questions, no retrieval, calibrated age-range overlap rubric). Source: arxiv.org. Saturation forecast: Estimated already saturated. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o Mini | 76.9 | #1627 |
| 2 | Claude 3 Haiku | 76.2 | #2629 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chronotqa-phenopackets-onset-no-retrieval · How It Works · Data refreshed daily, snapshot 2026-11-12.