HyperGVL - Isomorphism Recognition: leaderboard

HyperGVL (2026) hypergraph task for vision-language models: deciding whether two hypergraphs are isomorphic (yes/no), with each of 200 problems posed in 35 textual and visual representations. Accuracy (%); higher is better.

Metric: Accuracy (%) on deciding whether two hypergraphs are isomorphic (yes/no), 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 July 2027. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash64.6#175
2Qwen 2.5 VL 7B Instruct59.23#1902
3GPT-4o59.14#751
4GLM-4V-9B55.54#3142
5Gemma 3 12B51.29#2059
6InternVL2.5-8B51.06#2370
7InternVL3-8B50.54#1782
8Molmo-7B-D-092445.9#7207
9LLaVA-v1.5-7B44.53#7630
10Qwen 3 VL 8B42.79#917
11LLaVA-NeXT-7B42.27#5032
12DeepSeek-VL240.76#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-isomorphism-recognition · How It Works · Data refreshed daily, snapshot 2026-11-21.