GraphInstruct (Generation, Few-Shot CoT): leaderboard
Metric: Quality S_total (0-100; the paper's 0-1 score times 100): level scores weighted 0.05, 0.10, 0.15, 0.20, 0.25, 0.25 for L0 to L5, each level score a weighted mix of structural validity and distribution match, reference similarity and instruction-constraint satisfaction checked by code (no LLM judge; efficiency excluded), over the 800 instructions of GraphInstruct (instruction-following graph generation in six constraint levels); few-shot chain-of-thought prompting; five generations per instruction at temperature 0.7 with the InstructGraph code-style prefix; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 90.2 |
| 2 | Qwen 3.5 397B A17B | 87.9 |
| 3 | Qwen 3.5 122B A10B | 87.1 |
| 4 | Qwen 3.5 35B A3B | 86.1 |
| 5 | Llama 3.3 70B Instruct | 83.4 |
| 6 | GPT-4o | 82.3 |
| 7 | GPT-4.1 | 81.1 |
| 8 | GPT-4o Mini | 71.5 |
| 9 | GPT-3.5 Turbo | 71.3 |
| 10 | Llama 3.1 8B Instruct | 68.2 |
Interactive version: theaggregate.ai/benchmark?slug=graphinstruct-generation-few-shot-cot · How It Works · Data refreshed daily, snapshot 2026-10-07.