The authors formulate chart editing as cross-modal grounding: a model receives a rendered chart, its source program and a natural-language instruction, then modifies the code responsible for the requested visual change.
Chart editing benchmark separates completed requests from unwanted code changes
ChartRevise evaluates whether models make every requested and implied code change while preserving unrelated chart properties.
Big Tech
Jiaxiang Tang · Yi Zhou · Chad DeLuca · Rogerio Feris · Ahmed Khalil Omran · Zhi-Li Zhang · +2 more
University of Minnesota, Twin-Cities · IBM Research · Horizon School of Digital Technologies
Research Digest··2 min read
Tang et al.
Why this paper
From IBM Research and 2 others
In one line
Chart editing models need a dataset and evaluation protocol that separately measures requirement completion, coupled updates, and gratuitous changes.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§