IndoNutriQA - 5-shot F1: leaderboard
IndoNutriQA: 137 test questions from Indonesian clinical nutrition reports, split by source report. Identical Indonesian system prompt, temperature 0, fixed few-shot seed and no parameter updates. Exact Instruct, IT, medical and Indonesian checkpoints follow Table 2. Text overlap measures agreement with the reference answer. F1 (%).
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.1 8B Instruct | 55.55 | #1044 |
| 2 | MedGemma-4B-IT | 54.28 | #825 |
| 3 | GPT-4o | 50.33 | #351 |
| 4 | Llama 3 8B Cpt Sahabatai V1 Instruct | 43.32 | #917 |
| 5 | GPT-4o Mini | 43.07 | #604 |
| 6 | Qwen 2.5 7B Instruct | 42.62 | #870 |
No result here: #5 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=indonutriqa-5-shot-f1 · How It Works · Data refreshed daily, snapshot 2026-10-25.