Bio-VQA - Phylum: leaderboard
Fine-grained biological visual QA covering taxonomy, traits, spatial reasoning and individual re-identification. Taxonomic exact-match fraction. Bio-VQA v1 Table3 native headline columns,18 generalist checkpoints bound to Table13. Greedy decoding, no sampling/chain-of-thought/few-shot; one attempt per item, same prompts and scorer. Do not replace these Table3 measures with differently defined appendix headline metrics. Domain-specialist embedding baselines, Mean and Floor rows are not generic evaluated VLMs and are not admitted.
Metric: Taxonomic exact-match fraction. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3-38B-Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.91 | |
| 2 | Qwen 3 VL 8B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.89 | |
| 3 | Gemma 4 12B (IT) (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.88 | |
| 4 | Qwen 3 VL 30B A3B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.86 | |
| 5 | Qwen 3 VL 32B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.83 | |
| 6 | InternVL3-8B-Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.83 | |
| 7 | Molmo2-8B (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.82 | |
| 8 | InternVL3.5-8B-Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.82 | |
| 9 | Gemma 3 12B (IT) (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.79 | |
| 10 | Qwen 2.5 VL 7B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.76 |
Interactive version: theaggregate.ai/benchmark?slug=bio-vqa-phylum · How It Works · Data refreshed daily, snapshot 2026-11-06.