DV-World - DV-Evol Vega-Lite: leaderboard
Chart adaptation in Vega-Lite on DV-World (2026), run in the DV-World-Agent framework: the agent adapts a reference visualization to new data, scored half by an MLLM rubric against gold plots and half by coverage of the gold values. Percent; higher is better.
Metric: Overall score (%) on the Vega-Lite DV-Evol tasks (adapt a reference visualization to new data in the target library: Python Matplotlib/Seaborn, Apache ECharts, Vega-Lite, D3.js or Plotly.js); per-task score = 0.5 x MLLM rubric score (integrity, consistency, aesthetics against gold plots) + 0.5 x table coverage of gold values, 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 November 2027. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 50.68 | #319 |
| 2 | Gemini 3 Pro (Preview) | 46.3 | #119 |
| 3 | GPT-4.1 | 46.17 | #505 |
| 4 | Gemini 3 Flash | 45.39 | #175 |
| 5 | GPT-5.2 | 42.29 | #185 |
| 6 | Qwen 3 VL Plus | 28.02 | #512 |
| 7 | Gemini 2.5 Pro | 27.04 | #276 |
| 8 | Qwen 3 VL 32B | 26.21 | #622 |
| 9 | Qwen 3 VL 8B | 23.27 | #917 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=dv-world-dv-evol-vega-lite · How It Works · Data refreshed daily, snapshot 2026-11-21.