Writing study: Learning Visual Hierarchies in Hyperbolic Space for Image Retrieval

ICCV overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICCV 2025 oral writing corpus

Ziwei Wang, Sameera Ramasinghe, Chenchen Xu, Julien Monteil, Loris Bazzani, Thalaiyasingam Ajanthan · Official ICCV 2025 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

≈ 8.06 page-equivalents

Main figures / tables

6 / 4

Attached appendix figures / tables

0 / 0

Attached appendix span

No boundary detected; not assumed zero

Before references / appendix: ≈ 73.27%References / intervening material: ≈ 26.73%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	≈ 1.44 pp.	13.09%
2. Preliminaries	2	≈ 0.47 pp.	4.27%
3. Enforcing User-Defined Hierarchies	3	≈ 2.49 pp.	22.64%
4. Related Work	5	≈ 0.47 pp.	4.27%
5. Experiments	6	≈ 2.57 pp.	23.36%
6. Conclusion	8	≈ 0.25 pp.	2.27%
References	9	≈ 2.94 pp.	26.73%

All detected figure and table locations

Main text

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

Attached appendix

No recognized caption labels.

Argument and organization

1. 1 Introduction; 2 Preliminaries

   Motivates asymmetric visual hierarchy and introduces hyperbolic geometry before the learning rule.

2. 3 Enforcing User-Defined Hierarchies; 4 Related Work

   Defines containment-based pairs, bidirectional angle loss and a distribution-sensitive retrieval metric.

3. 5 Experiments; 6 Conclusion

   Separates same-class from hierarchical retrieval, cross-image sampling effects and out-of-domain tests.

Problem, claim and contribution

The no-hierarchical-label claim is paired with an explicit source of supervision: object boxes/classes generate containment and cross-image entailment pairs.

Sections 1 and 3.1, physical pp.1–4 · Main paper

Argument and section order

The evaluation metric is defined alongside the hierarchy construction and loss, before numerical comparisons, clarifying why ordinary recall alone is insufficient.

Sections 3.1–3.3, pp.3–5 · Main paper

Figures, captions and visual explanation

A three-panel example links nested boxes to a hierarchy and then to same-class boxes from other images, visually distinguishing two sources of positive pairs.

Figure 2, p.4 · Main paper

Tables, comparisons and evidence

Separate tables report same-class retrieval and hierarchy-distribution alignment; the precision–recall plot labels decision thresholds and uses line style to distinguish cross-image supervision.

Tables 1–2 and Figure 4, p.7 · Main paper

Appendix and reproducibility

The emergent high-frequency-crop example directs further detail to supplementary materials; this scan inspected only the static main examples and pointer.

Figure 6 and Section 5.2, p.8 · Main paper

Clarity and writing conventions

The downstream detection caption explicitly says ground-truth boxes and top-five majority voting, preventing a region-classification experiment from silently reading as end-to-end detection.

Table 4 and Section 5.2, p.8 · Main paper

What to reuse—and what not to infer

Transferable lesson: Define the hierarchy and asymmetric retrieval target operationally, and expose privileged regions in downstream evaluation captions.

Boundary: No explicit hierarchy tree during training does not mean no annotation supervision. Box containment and class matching encode a chosen hierarchy rather than proving universal human semantic organization; geometric arguments were not proof-verified.

Apply this lesson using the argument and evidence worksheet.

Inspection record and sources

Checked

2026-10-07

Reading status

Writing-focused manuscript scan + selected visuals inspected

Main text

Writing-oriented scan across the main manuscript, physical PDF pp.1–8, covering framing, section progression, method explanations, captions, comparison protocols and concluding scope; reference boundary checked. Selected full-page rasters were inspected. Not exhaustive line-by-line reading, proof verification or experimental reproduction.

Appendix / supplement

Main-paper appendix/supplement references were checked. A separate supplementary source is linked in the inventory but was not retrieved or inspected; its contents are not included in the reading or visual-page counts.

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 CVF open-access ICCV 2025 accepted-paper PDF, retrieved 7 October 2026; CVF states identical to accepted version except watermark; not assumed identical to initial submission · 11 pages