RealBirdID - Unanswerable Genus Accuracy: leaderboard
Metric: Genus-level accuracy (%) on the unanswerable RealBirdID images (observations the iNaturalist community could only resolve to genus, because the species needs song, a hidden view or a sharper image), zero-shot, each model asked 'What is the species of this bird?' about one wild bird photo from iNaturalist and its free-form answer mapped to the taxonomy by constrained decoding (nlg2choice), micro-averaged; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 60.1 | #145 |
| 2 | GPT-5 | 58.6 | #91 |
| 3 | Qwen 2.5 VL 7B Instruct | 52.6 | #643 |
| 4 | Llama 3.2 11B Instruct | 42.7 | #1112 |
| 5 | Gemma 3 12B (IT) | 37.4 | #655 |
| 6 | InternVL3-8B | 34.7 | #606 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=realbirdid-unanswerable-genus-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-11.