DV-World - DV-Sheet Fix: leaderboard
Metric: Chart repair success rate (%): share of diagnostic spreadsheet tasks whose repaired native chart matches the healthy original on every must-fix attribute (similarity at least 0.95), run in the DV-World-Agent framework (data manipulation, multimodal perception and proactive interaction tools); expert rubrics applied by a Gemini-2.5-Flash judge where rubric-scored, means over evaluation trials; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 48 |
| 2 | Kimi K2 (Thinking) | 46 |
| 3 | GPT-5.2 | 42 |
| 4 | Grok 4 | 42 |
| 5 | Gemini 2.5 Pro | 40 |
| 6 | GLM-4.7 | 40 |
| 7 | DeepSeek V3.2 | 36 |
| 8 | GPT-4.1 | 32 |
| 9 | Qwen 3 Coder Plus | 24 |
| 10 | Qwen 3 235B A22B | 22 |
| 11 | Qwen 3 8B | 14 |
Interactive version: theaggregate.ai/benchmark?slug=dv-world-dv-sheet-fix · How It Works · Data refreshed daily, snapshot 2026-10-07.