SDABench (MCQ): leaderboard
The multiple-choice format of SDABench (Southeast University, ECNU and HKUST, 2026): scientific data-analysis questions on held-out synthetic datasets from five domains, each with four options including perturbation-based distractors. The score is accuracy in percent averaged over six capability types, with 25 percent at chance; higher is better.
Metric: Accuracy (%) of four-option multiple-choice answers with perturbation-based distractors on the held-out SDA-Synth test split (2,400 instances, macro-averaged over the six task types; zero-shot; 25% chance); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 69.46 | #94 |
| 2 | GPT-5.4 | 66.58 | #123 |
| 3 | Claude Sonnet 4.6 | 66.38 | #161 |
| 4 | Qwen 3.5 397B A17B | 62.46 | #246 |
| 5 | DeepSeek V3.2 | 62.21 | #406 |
| 6 | GLM-5 | 60.79 | #260 |
| 7 | GPT-5.4 Mini | 58.46 | #389 |
| 8 | Kimi K2.5 | 57.04 | #256 |
| 9 | DeepSeek R1 | 56.38 | #473 |
| 10 | Qwen 3 235B A22B 2507 Instruct | 52.42 | #619 |
| 11 | Llama 3.3 70B Instruct | 30.46 | #1357 |
| 12 | Llama 3.1 8B Instruct | 25 | #4246 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sdabench-mcq · How It Works · Data refreshed daily, snapshot 2026-11-13.