Making Multimodal LLMs Reliable Chart Data Extractors: A Benchmark and Training Framework

A lab notebook opened to its best page: a framed video like a monitor bezel, emoji in the headline, and every figure pinned with a soft shadow.

ExChart.github.ioMaking Multimodal LLMs Reliable Chart Data Extractors: A Benchmark and Training Framework
Screenshot of https://ExChart.github.io/, © the authors. Shown for reference. Request removal
Venue
CHI 2026
Year
2026
First author
Yuchen He
Field
HCI / visualization, MLLMs
Template
Nerfies (Bulma 0.9)
Theme
Light

Checked against the rendered page. Only Google Sans 400 loads, so the 600 title is synthetic bold; the video's first frame shows a loading spinner ring.

Why it is here

Rich, well-framed presentation of the work: a framed video teaser, a TL;DR, a benchmark pipeline figure and a key-features list with a colorful results panel. Centered layout with good hierarchy.

Overview

Color Palette

Typography

Google Sans
Making Multimodal LLMs Reliable Chart Data Extractors: A Benchmark and Abstract · Method · Results · BibTeX — 0123456789
Noto Sans
Making Multimodal LLMs Reliable Chart Data Extractors: A Benchmark and Abstract · Method · Results · BibTeX — 0123456789

Spacing & Shape

Page Modules

Guidelines

More like this