PolyChartQA - MLM-Generated Questions: leaderboard

Question answering over multi-chart images on PolyChartQA (2026, Beyond Single Plots), machine-generated portion: 2,175 questions written by GPT-4.1 and manually verified, about figures combining several related charts from computer science papers. Zero-shot prompting; accuracy in percent rated by an LLM judge, higher is better.

Metric: L-Accuracy (%, times 100), zero-shot prompting, on the 2,175 GPT-4.1-generated, manually verified questions of PolyChartQA (534 multi-chart images with 2,297 sub-charts from computer science papers), answers rated correct by a Claude 3.7 Sonnet judge (L-Accuracy); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.0 Flash80.28#723
2Claude 3.7 Sonnet74.94#470
3Pixtral-12B63.77#2553
4ChartGemma28.28#8644
5Gemma 3 4B (IT)27.49#4139
6Llama 3.2 11B Vision17.94#4264
7LLaVA-v1.5-7B11.68#7630
8MatCha5.43

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=polychartqa-mlm-generated-questions · How It Works · Data refreshed daily, snapshot 2026-11-21.