EstGraph - Node Count, GRP Medium (MH Walk): leaderboard
Estimating how many nodes a Gaussian random partition graph of 1,000-10,000 nodes has from Metropolis-Hastings random-walk statistics that cover 20% of it; EstGraph (Maurya and Liu, 2026). Median relative error of the estimate in percent over 100 synthetic graphs; lower is better.
Metric: Median relative error of the node-count estimate (%; 100 synthetic Gaussian random partition graphs of 1,000-10,000 nodes, the model reads Metropolis-Hastings random-walk statistics covering 20% of the graph). Source: arxiv.org. Saturation forecast: Around December 2026. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 (High) | 3.09 | #224 (O3) |
| 2 | DeepSeek V3.1 (Thinking) | 3.52 | #500 (DeepSeek V3.1) |
| 3 | Claude Sonnet 4 | 11.2 | #397 |
| 4 | Gemini 2.5 Pro | 15.37 | #262 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=estgraph-node-count-grp-medium-mh-walk · How It Works · Data refreshed daily, snapshot 2026-11-12.