HyperGVL - Order-Filtered Neighbors: leaderboard

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

Metric: Accuracy (%) on listing a vertex's neighbors through hyperedges of at least 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: Estimated already saturated. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash97.13#180
2GPT-4o82.86#754
3Gemma 3 12B50.87#2091
4Qwen 3 VL 8B43.13#927
5InternVL3-8B30.54#1795
6Qwen 2.5 VL 7B Instruct27.17#1919
7InternVL2.5-8B22#2393
8GLM-4V-9B18.9#3164
9Molmo-7B-D-092414.4#7206
10LLaVA-NeXT-7B8.39#5045
11DeepSeek-VL25.11#3056
12LLaVA-v1.5-7B1.61#7642

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

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