ClosureBench (Chain-of-Thought) - L2 Chain: leaderboard
Metric: Accuracy (%; lenient scoring of one JSON answer per question; L2 chained queries (two to three steps, 400 instances); held-out split of 1,300 generated graph questions over graph sizes 4-10, temperature 0; chain-of-thought mode: reason step by step, then answer). Source: arxiv.org. Saturation forecast: Estimated already saturated. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4 | 90.2 |
| 2 | GPT-4.1 | 89 |
| 3 | GPT-4.1 Mini | 88 |
| 4 | Gemini 2.5 Flash | 82.8 |
Interactive version: theaggregate.ai/benchmark?slug=closurebench-chain-of-thought-l2-chain · How It Works · Data refreshed daily, snapshot 2026-09-26.