SciDiagramEdit: Learning to Edit Scientific Diagrams from Paper Revisions

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Yasheng Sun, Zezi Zeng, Yifan Yang, Chong Luo, Wenyi Wang, Ziwei Liu, Jürgen Schmidhuber
• Venue: arXiv: cs.CL
• Published: 2026-07-16
• Collected: 2026-07-17
• URL: http://arxiv.org/abs/2607.15272v1
• PDF: https://arxiv.org/pdf/2607.15272v1
• DOI: UNKNOWN

Plain-Language Brief

The text introduces SciDiagramEdit, a framework for automating the editing of scientific diagrams in research papers based on natural language instructions. The framework learns from natural paper revisions and operates on the editable vector source of figures, allowing users to inspect and co-edit individual primitives alongside the agent.

What the Document Says

• The text discusses SciDiagramEdit, a tool for editing scientific diagrams in research papers based on natural language instructions.
• SciDiagramEdit learns from natural paper revisions and operates on the editable vector source of figures.
• Users can inspect and co-edit individual primitives alongside the agent.

Key Concepts Mentioned

• Scientific diagram editing
• Natural language instructions
• Vector source
• Skill evolution

Claims Made by the Document

• SciDiagramEdit learns from natural paper revisions.
• Users can inspect and co-edit individual primitives alongside the agent.

Evidence Used

• The text states that SciDiagramEdit operates on the figure's editable vector source, where users can inspect and co-edit individual primitives alongside the agent.
• The text mentions that the framework learns from natural paper revisions.

Methods / Approach

• Agentic learning via skill evolution
• Mining before/after figure pairs from arXiv version histories

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Editing the figures in a research paper is a routine and time-consuming part of everyday research practice: authors relabel components, rearrange panels, and restyle visuals as they revise their manuscripts. Automating this editing workflow under a natural-language instruction, however, is challenging, because a scientific figure is a dense infographic in which heterogeneous visual elements such as schematics, plots, photos, captions, and arrows are composed under a tight visual grammar to advance a specific argument. To address this, we present SciDiagramEdit, a benchmark and skill-evolution framework that learns from natural paper revisions and operates on the figure's editable vector source, where users can inspect and co-edit individual primitives alongside the agent. Our benchmark mines before/after figure pairs from arXiv version histories, each grounded in the authors' own revision intent. To accommodate the diversity of editing instructions, we adopt agentic learning via skill evolution: an agentic proposer continually refines the agent's skill specification from execution traces over multiple epochs. The resulting skill progressively lifts edit accuracy on a held-out validation set, providing evidence that natural paper revisions are an effective training signal for instruction-driven figure editing.