HyperGVL - 3-Coloring: leaderboard

HyperGVL (2026) hypergraph task for vision-language models: giving a valid 3-coloring in which every hyperedge has at least two colors, with each of 200 problems posed in 35 textual and visual representations. Accuracy (%); higher is better.

Metric: Accuracy (%) on giving a valid 3-coloring in which every hyperedge has at least two colors, 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 Flash79.34#180
2Qwen 3 VL 8B57.07#927
3GPT-4o56.86#754
4InternVL3-8B38.04#1795
5Gemma 3 12B33.69#2091
6Qwen 2.5 VL 7B Instruct26.31#1919
7InternVL2.5-8B22.76#2393
8GLM-4V-9B20.57#3164
9LLaVA-NeXT-7B15.84#5045
10LLaVA-v1.5-7B10.04#7642
11Molmo-7B-D-09246.63#7206
12DeepSeek-VL21.4#3056

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=hypergvl-3-coloring · How It Works · Data refreshed daily, snapshot 2026-11-22.