DRAGON : A Benchmark for Evidence-Grounded Visual Reasoning over Diagrams
Anirudh Iyengar Kaniyar Narayana Iyengar, Tampu Ravi Kumar Gaurav Najpande , Manan Suri , Dinesh Manocha , Puneet Mathur , Vivek Gupta [2pt] Arizona State University Adobe Research University of Maryland[2pt] [email protected], [email protected], [email protected], [email protected], [-4pt] [email protected], [email protected], [email protected] · 2026
Read DRAGON : A Benchmark for Evidence-Grounded Visual Reasoning over Diagrams by Anirudh Iyengar Kaniyar Narayana Iyengar, Tampu Ravi Kumar Gaurav Najpande , Manan Suri , Dinesh Manocha , Puneet Mathur , Vivek Gupta [2pt] Arizona State University Adobe Research University of Maryland[2pt] [email protected], [email protected], [email protected], [email protected], [-4pt] [email protected], [email protected], [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-25231.txt.
Description
Iyengar and six colleagues introduce DRAGON, a benchmark for evidence-grounded visual reasoning over diagrams. Diagram question answering needs models to interpret charts and maps, yet correct answers do not guarantee grounding. Given a diagram, a question and the answer, a model must predict bounding boxes for the elements justifying the answer. The data set holds 11,664 instances from six diagram data sets, with human-verified evidence annotations. Even Claude Opus 4.6 reaches a Grounding IoU of only 23.8 percent on AI2D, so faithful grounding remains an open challenge.
Catalog record
| CREATOR | Anirudh Iyengar Kaniyar Narayana Iyengar, Tampu Ravi Kumar Gaurav Najpande , Manan Suri , Dinesh Manocha , Puneet Mathur , Vivek Gupta [2pt] Arizona State University Adobe Research University of Maryland[2pt] [email protected], [email protected], [email protected], [email protected], [-4pt] [email protected], [email protected], [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71039 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.25231