Writing study: CURE: Curriculum-guided Multi-task Training for Reliable Anatomy Grounded Report Generation

CVPR overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← CVPR 2026 oral writing corpus

Pablo Messina, Andrés Villa, Juan León Alcázar, Karen Sanchez, Carlos Hinojosa, Denis Parra, Alvaro Soto, Bernard Ghanem · Official CVPR 2026 oral record

Observed writing choices, not acceptance rules. This study examines the inspected manuscript version. It does not establish why the paper was accepted or that its presentation choices should be copied unchanged.

Figure, table and section ledger

Editorial contribution type: Training, objectives & optimization (primary).

Layout extraction, not a required paper template. Counts are distinct caption labels; a multi-panel figure counts once. Section spans are estimated from heading positions, including floats and whitespace. Method and limitations.

Canonical PDF

11 physical pages

Before references / appendix

≈ 8.14 page-equivalents

Main figures / tables

3 / 7

Attached appendix figures / tables

0 / 0

Attached appendix span

Separate file recorded below; not in this PDF total

Before references / appendix: ≈ 74%References / intervening material: ≈ 26%Estimated layout envelopes, not text/artwork area. Unresolved appendix boundaries remain within the undivided post-body material.

Extraction notes: No attached appendix boundary detected. Zero means no recognized caption, not independently verified absence.

Section-by-section lengths and source pages

The title/abstract/front matter precede the first numbered heading. A span ends at the next detected top-level heading, not at the last sentence; end-page whitespace is included. References and statements are listed separately. These are not measurements of text-only or artwork area.

Top-level heading	Starts on PDF page	Estimated span	Share of entire PDF
1. Introduction	1	≈ 0.93 pp.	8.45%
2. Related Work	2	≈ 0.6 pp.	5.45%
3. Methodology	2	≈ 2.5 pp.	22.73%
4. Experiments	5	≈ 3.04 pp.	27.64%
5. Conclusion	8	≈ 0.5 pp.	4.55%
Acknowledgments	8	≈ 0.22 pp.	2%
References	9	≈ 2.86 pp.	26%

All detected figure and table locations

Main text

• Figure 1 · p. 1
• Table 1 · p. 3
• Figure 2 · p. 4
• Table 2 · p. 5
• Table 3 · p. 6
• Table 4 · p. 7
• Table 5 · p. 7
• Table 6 · p. 7
• Table 7 · p. 8
• Figure 3 · p. 8

Attached appendix

No recognized caption labels.

Separate source files—excluded from canonical-PDF totals

A separate author version may repeat main text and references. Whole-file lengths and counts are not extra appendix material and are never added to the canonical PDF totals. Read the study's source scope for which sections received substantive reading versus layout inspection.

Additional source 1

Separate PDF (or ZIP containing it): whole-file layout of 20 physical pages; 4 detected figures and 13 tables. These may include overlapping main-text material.

Located appendix only: starts on physical p.1; ≈ 19.85 page-equivalents, 4 figure labels and 13 table labels. This interval is distinct from the whole file.

Detected heading	PDF page	Estimated span
Supplementary Material	1	≈ 0.04 pp.
6. Implementation and Training Details	1	≈ 1.36 pp.
7. Datasets and Task Formulation	2	≈ 3.02 pp.
8. Evaluation Protocol	5	≈ 0.63 pp.
9. Detailed Experimental Results	6	≈ 8.9 pp.
10. Additional Qualitative Analysis	15	≈ 5.91 pp.

• Table 8 · p. 1
• Figure 4 · p. 2
• Figure 5 · p. 2
• Table 9 · p. 4
• Table 10 · p. 7
• Table 11 · p. 7
• Table 12 · p. 8
• Table 13 · p. 10
• Table 14 · p. 10
• Table 15 · p. 12
• Table 16 · p. 12
• Table 17 · p. 13
• Table 18 · p. 16
• Table 19 · p. 17
• Table 20 · p. 18
• Figure 6 · p. 19
• Figure 7 · p. 20

Argument and organization

1. 3. Methodology

   Defines curriculum updates and grounded task formats.

2. 4. Experiments

   Separates grounding, report generation, hallucination, ablation, and qualitative evidence.

Problem, claim and contribution

A false-positive fracture example motivates factuality separately from locating the relevant anatomy.

Figure 1 p. 1 · Main paper · Additional source 2

Argument and section order

Dataset/task/split accounting precedes the curriculum details, helping readers distinguish training data from evaluation-only datasets.

Table 1 p. 3; §3 · Main paper · Additional source 2

Figures, captions and visual explanation

The overview aligns a feedback-driven sampler with its constituent datasets and three output tasks; abbreviations are expanded within the panel.

Figure 2 p. 4 · Main paper · Additional source 2

Tables, comparisons and evidence

Task-specific tables preserve trade-offs: better grounding need not imply better text metrics, and the prose acknowledges MAIRA-2's text advantages.

§4.1 p. 6; Tables 4–6 p. 7 · Main paper · Additional source 2

Appendix and reproducibility

Extended ablations pair configuration definitions with result rows; the hallucination audit discloses prompts and an automated judge rather than treating it as direct clinical adjudication.

Supplement Tables 10–11 and §9.4 · Main paper · Additional source 2

Clarity and writing conventions

Training, zero-shot testing, and proprietary baseline data are qualified explicitly; these conditions matter more than a single overall superiority sentence.

Figure 2 caption; §4.1 · Main paper · Additional source 2

What to reuse—and what not to infer

Transferable lesson: For multitask medical generation, map each claim to its own dataset, output contract, metric, and evaluation protocol.

Boundary: Automated report and NLI scores are not clinical safety validation; improvements are not uniform across anatomy or metric.

Apply this lesson using the argument and evidence worksheet.

Inspection record and sources

Checked

2026-09-26

Reading status

Writing-focused manuscript scan + selected visuals inspected

Main text

Manuscript-wide writing scan of main PDF pp. 1–8: framing, section structure, method exposition, figure/table captions, evaluation organization and conclusions; not an exhaustive technical proof audit or reproduction.

Appendix / supplement

Scanned 20 supplementary pages spanning hyperparameters, task I/O and splits, curriculum dynamics, variant definitions, full metrics, automated hallucination/NLI judging, and additional examples.

Visually inspected PDF pages

4, 7

This is a writing-focused scan of the manuscript and the stated supplementary scope: framing, organization, visual explanation, evidence presentation and wording. It is not a line-by-line proof audit, reproduction of results or exhaustive scientific review.

• Official CVPR 2026 CVF proceedings PDF; retrieved 2026-09-26 · 11 pages
• Official CVPR 2026 separate supplementary PDF; retrieved 2026-09-26 · 20 pages