ChartFI - Informativeness: leaderboard
ChartFI (2026) tests how faithfully and insightfully multimodal LLMs describe complex scientific charts, over 896 charts from peer-reviewed arXiv papers with author-grounded reference descriptions. This board is informativeness, from 0 to 1: how much of the generated description consists of higher-level perceptual and domain insights rather than chart construction details, weighted against the reference description; higher is better.
Metric: Informativeness (0-1): mean semantic-level weight of the generated insights, weighting higher-level perceptual and domain insights more, with weights adapted to the reference description, descriptions of the 896 ChartFI-Bench scientific charts (arXiv papers accepted at peer-reviewed venues), temperature 0 and top-p 1; higher is better. Source: arxiv.org. Saturation forecast: Around 2034. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 0.3188 | #138 |
| 2 | Llama 4 Scout | 0.3058 | #1825 |
| 3 | Claude Sonnet 4.6 | 0.3033 | #179 |
| 4 | Qwen 3.5 Plus | 0.2854 | #250 |
| 5 | Gemini 3 Flash | 0.2762 | #180 |
| 6 | Qwen 3.5 27B | 0.2747 | #380 |
| 7 | InternVL3.5-14B | 0.2667 | #1519 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=chartfi-informativeness · How It Works · Data refreshed daily, snapshot 2026-11-22.