HyperGVL - Strict Hypercycle: leaderboard

HyperGVL (2026) hypergraph task for vision-language models: finding a strict hypercycle, adjacent hyperedges sharing exactly one vertex, with each of 200 problems posed in 35 textual and visual representations. Accuracy (%); higher is better.

Metric: Accuracy (%) on finding a strict hypercycle, adjacent hyperedges sharing exactly one vertex, 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 Flash66.44#175
2Qwen 3 VL 8B46.57#917
3GPT-4o25.14#751
4GLM-4V-9B14.19#3142
5Gemma 3 12B13.39#2059
6Qwen 2.5 VL 7B Instruct10.51#1902
7LLaVA-v1.5-7B8.44#7630
8Molmo-7B-D-09246.8#7207
9InternVL2.5-8B5.94#2370
10InternVL3-8B5.24#1782
11LLaVA-NeXT-7B2.83#5032
12DeepSeek-VL20.43#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-strict-hypercycle · How It Works · Data refreshed daily, snapshot 2026-11-21.