HyperGVL - Order-Specified Hyperedge Counting: leaderboard

HyperGVL (2026) hypergraph task for vision-language models: counting hyperedges with a given order, with each of 200 problems posed in 35 textual and visual representations. Accuracy (%); higher is better.

Metric: Accuracy (%) on counting hyperedges with a given order, 7,000 questions per task (200 problems on synthetic and real-world hypergraphs, each posed in 35 textual, visual and combined hypergraph representations), accuracy (%) averaged over the representations, temperature 0.8, top-p 0.95; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash77.83#175
2GPT-4o50.57#751
3Qwen 3 VL 8B49.14#917
4Gemma 3 12B47.35#2059
5InternVL3-8B35.71#1782
6Qwen 2.5 VL 7B Instruct29.01#1902
7GLM-4V-9B23.33#3142
8InternVL2.5-8B18.26#2370
9LLaVA-v1.5-7B16.57#7630
10Molmo-7B-D-092415.81#7207
11LLaVA-NeXT-7B14.59#5032
12DeepSeek-VL210.6#2927

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=hypergvl-order-specified-hyperedge-counting · How It Works · Data refreshed daily, snapshot 2026-11-21.