figure-caption-writer

Figure Caption Writer

Purpose

Write figures and captions that communicate evidence clearly. A good figure should make a claim easier to evaluate, not merely decorate the paper.

When To Use

• Designing the paper's figure plan.
• Revising captions for clarity and reviewer usefulness.
• Checking whether figures support claims in Abstract, Introduction, or Results.
• Preparing CVPR/ECCV or TMI submissions where visual evidence is central.

Inputs

Required:

• figure image or description;
• claim the figure supports;
• target venue;
• related section text;
• experimental setting or data source.

Optional:

• panel labels;
• metric definitions;
• qualitative examples;
• failure cases;
• accessibility constraints.

Procedure

1. Identify the figure's purpose: overview, method diagram, quantitative result, qualitative result, failure case, dataset example, or ablation.
2. State the takeaway in the caption's first sentence.
3. Define panels, colors, symbols, and metrics.
4. Connect the figure to a claim in the paper.
5. Avoid hiding important experimental conditions in tiny text.
6. Check anonymity and data/privacy constraints.
7. Ensure the main text calls out the figure.

Rubric

Dimension	Strong	Weak
Takeaway	Caption states what to learn	Caption only names components
Self-contained	Panels/metrics clear	Reader must guess
Evidence link	Supports a paper claim	Decorative
Visual clarity	Readable and focused	Too dense or tiny
Compliance	Anonymous/privacy-safe	Leaks identity or sensitive data

Venue Adaptation

• ICLR/NeurIPS/ICML: use figures to explain mechanisms and summarize evidence compactly.
• CVPR/ECCV: qualitative figures and visual comparisons must be fair, readable, and not cherry-picked.
• AAAI: keep captions understandable to broad AI readers.
• TMI: protect patient privacy; include modality, view, and relevant clinical context when appropriate.
• arXiv: ensure public images have permissions and no anonymous-review artifacts.