HyperGVL - Reasoning: leaderboard

HyperGVL (2026) tests vision-language models on hypergraph reasoning: order-weighted shortest path and maximum flow, isomorphism, 3-coloring, strict hypercycles and Hamilton paths, each problem posed in 35 textual and visual representations. Average accuracy (%) over the six reasoning tasks; higher is better.

Metric: Average accuracy (%) over the six reasoning tasks (order-weighted shortest path and maximum flow, isomorphism, 3-coloring, strict hypercycle, Hamilton path), 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 Flash62.04#175
2GPT-4o35#751
3Qwen 3 VL 8B34.43#917
4Gemma 3 12B25.53#2059
5Qwen 2.5 VL 7B Instruct21.3#1902
6InternVL3-8B20.52#1782
7GLM-4V-9B18.9#3142
8InternVL2.5-8B16.35#2370
9Molmo-7B-D-092412.83#7207
10LLaVA-v1.5-7B12.28#7630
11LLaVA-NeXT-7B11.99#5032
12DeepSeek-VL28.21#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-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-21.