Bio-VQA - Anatomical Naming: leaderboard
Fine-grained biological visual QA covering taxonomy, traits, spatial reasoning and individual re-identification. Chance-adjusted graded score (cags). 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: Chance-adjusted graded score (cags). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 32B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.73 | |
| 2 | Qwen 2.5 VL 7B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.71 | |
| 3 | InternVL3-38B-Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.67 | |
| 4 | Qwen 2.5 VL 3B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.66 | |
| 5 | Qwen 3 VL 30B A3B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.66 | |
| 6 | Qwen 3 VL 8B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.64 | |
| 7 | InternVL3-8B-Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.58 | |
| 8 | InternVL3.5-30B-A3B-Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.54 | |
| 9 | InternVL3.5-38B-Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.54 | |
| 10 | Llama 3.2 11B Instruct (Benchmark Sweep Run: source=BioVQA; greedy; zero-shot; noCoT; single-attempt; exact Table13 checkpoint) | 0.52 |
Interactive version: theaggregate.ai/benchmark?slug=bio-vqa-anatomical-naming · How It Works · Data refreshed daily, snapshot 2026-11-06.