DSGym - QRData-Verified: leaderboard
Statistical and causal reasoning questions over tabular data from QRData, re-verified by DSGym to drop invalid or ambiguous items and questions that models answer without reading the data, run by DSGym's default CodeAct agent (Nie et al., 2026) in an isolated Jupyter container at temperature 0 with all tools disabled. Exact-match accuracy with a small numeric tolerance, in percent; higher is better.
Metric: Exact-Match Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2 0905 | 63.68 | #555 |
| 2 | GPT-5 (Medium) | 61.75 | #163 (GPT-5) |
| 3 | Claude Sonnet 4.5 | 61.35 | #244 |
| 4 | GPT-4o (2024-08-06) | 60.24 | #731 |
| 5 | GPT-5.1 (High) | 60.16 | #225 (GPT-5.1) |
| 6 | Claude Sonnet 4 | 59.06 | #397 |
| 7 | GPT-5.1 (Non-reasoning) | 58.96 | #225 (GPT-5.1) |
| 8 | DeepSeek V3.1 | 57.37 | #500 |
| 9 | GPT-OSS-120B | 47.95 | #699 |
| 10 | Qwen 3 4B Instruct | 45.27 | #1660 |
| 11 | Qwen 2.5 7B Instruct | 35.04 | #2583 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=dsgym-qrdata-verified · How It Works · Data refreshed daily, snapshot 2026-11-12.