Yuchen He @y_he02 - Twitter Profile

Yuchen He @y_he02

4 months ago

🔗 https://t.co/B08POHebBM

0

78

Yuchen He @y_he02

4 months ago

Chart data extraction is still tedious or unreliable. We study MLLMs and find a key gap: they get the structure right but miss the numbers. We introduce a benchmark + human-inspired training approach, and explore mixed-initiative workflows. Come talk to us at #CHI2026!

y_he02's tweet photo. Chart data extraction is still tedious or unreliable.
We study MLLMs and find a key gap: they get the structure right but miss the numbers.
We introduce a benchmark + human-inspired training approach, and explore mixed-initiative workflows.
Come talk to us at #CHI2026! https://t.co/qxAnHXXLTI

1

4

0

1

219

y_he02 retweeted

Dazhen @Dakzen4

10 months ago

Just presented our paper ReSpark at #UIST2025! ✨ ReSpark leverages previous data reports to help LLMs generate new analysis code, charts, and narratives—making data report authoring more efficient and logical. Code is open source 👉 https://t.co/KqHF9hNzS9 @zjuidg @bebincaa

Dakzen4's tweet photo. Just presented our paper ReSpark at #UIST2025! ✨

ReSpark leverages previous data reports to help LLMs generate new analysis code, charts, and narratives—making data report authoring more efficient and logical.

Code is open source 👉 https://t.co/KqHF9hNzS9

@zjuidg @bebincaa https://t.co/5wSrIXE3j8

0

2

0

136

Yuchen He @y_he02

over 1 year ago

@minsunkimhappy On the same session! Very interesting presentation!

1

0

97

Yuchen He @y_he02

over 1 year ago

Amazing first #chi2025! Check our paper about video data programming, which helps users define labeling functions to cold-start TAL model training: “ProTAL: A Drag-and-Link Video Programming Framework for Temporal Action Localization.” https://t.co/6NkespcQgB