Question Answering on ChronoTQA Tier 1 (147 Phenopackets-grounded onset questions)
86.4AccuracyDeepSeek V3
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeepSeek V3Retrieval Condition=NR2026.05 | 86.4 | — | |
| DeepSeek V3Retrieval Condition=ChronoMedKG2026.05 | 85 | 1.4 | |
| DeepSeek V3Retrieval Condition=HPOA2026.05 | 83.7 | — | |
| Pooled (n=441)Retrieval Condition=ChronoMedKG2026.05 | 82.1 | 2.3 | |
| Claude 3 HaikuRetrieval Condition=ChronoMedKG2026.05 | 81.6 | 5.4 | |
| DeepSeek V3Retrieval Condition=PrimeKG2026.05 | 81.6 | — | |
| GPT-4o-miniRetrieval Condition=PrimeKG2026.05 | 81.6 | — | |
| GPT-4o-miniRetrieval Condition=HPOA2026.05 | 81 | — | |
| Pooled (n=441)Retrieval Condition=HPOA2026.05 | 80.1 | — | |
| Pooled (n=441)Retrieval Condition=NR2026.05 | 79.8 | — | |
| GPT-4o-miniRetrieval Condition=ChronoMedKG2026.05 | 79.6 | 2.7 | |
| Pooled (n=441)Retrieval Condition=PrimeKG2026.05 | 77.3 | — | |
| GPT-4o-miniRetrieval Condition=NR2026.05 | 76.9 | — | |
| Claude 3 HaikuRetrieval Condition=NR2026.05 | 76.2 | — | |
| Claude 3 HaikuRetrieval Condition=HPOA2026.05 | 75.5 | — | |
| Claude 3 HaikuRetrieval Condition=PrimeKG2026.05 | 68.7 | — |