Writing study: Heavy Labels Out! Dataset Distillation with Label Space Lightening

ICCV overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICCV 2025 oral writing corpus

Ruonan Yu, Songhua Liu, Zigeng Chen, Jingwen Ye, Xinchao Wang · 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: Efficiency, inference & systems (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

10 physical pages

Before references / appendix

≈ 8.09 page-equivalents

Main figures / tables

3 / 7

Attached appendix figures / tables

0 / 0

Attached appendix span

No boundary detected; not assumed zero

Before references / appendix: ≈ 80.9%References / intervening material: ≈ 19.1%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	≈ 1.35 pp.	13.5%
2. Related Works	2	≈ 0.52 pp.	5.2%
3. Methods	3	≈ 2.07 pp.	20.7%
4. Experiments	5	≈ 3.49 pp.	34.9%
5. Conclusion	8	≈ 0.25 pp.	2.5%
Acknowledgments	9	≈ 0.04 pp.	0.4%
References	9	≈ 1.91 pp.	19.1%

All detected figure and table locations

Main text

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

Attached appendix

No recognized caption labels.

Argument and organization

1. 1 Introduction; 2 Related Works

   Identifies augmented soft-label storage as a hidden cost of apparently tiny distilled datasets.

2. 3 Methods

   Defines label expansion, text-initialized projection, low-rank transfer and synthetic-image updates before an algorithm summary.

3. 4 Experiments; 5 Conclusion

   Separates baseline/architecture transfer, component and rank controls, quantized labels, aerial imagery and continual learning.

Problem, claim and contribution

Turns the size of the label representation, not just images per class, into the main compression contract.

Abstract and Section 1, physical pp.1–2 · Main paper

Argument and section order

A short equivalence proposition connects text embeddings to the projector initialization; the algorithm then distinguishes offline transfer from online relabeling.

Sections 3.2–3.5 and Algorithm 1, pp.3–5 · Main paper

Figures, captions and visual explanation

The before/after diagram shows stored augmented labels replaced by an online projector, with frozen public features separated from the small learned matrices.

Figure 1, p.2 · Main paper

Tables, comparisons and evidence

Tables identify reproduced baselines, IPC settings, downstream architectures and what extra-storage rows include; high-IPC non-wins remain in the primary comparison.

Tables 1–2, p.6 · Main paper

Appendix and reproducibility

Detailed experimental settings are assigned to supplementary materials in the setup; this scan checked the main algorithm and evaluation but not those external details.

Section 4.1, p.5 · Main paper

Clarity and writing conventions

The text repeatedly calls the matrices extra storage, but headline compression rhetoric needs that qualifier because an available pretrained encoder still must execute.

Figure 1 caption, p.2; Tables 1 and 6, pp.6,8 · Main paper

What to reuse—and what not to infer

Transferable lesson: Audit all auxiliary assets behind a compact dataset claim and state whether storage is total, incremental or amortized over shared models.

Boundary: The tiny label-storage ratio excludes the shared frozen foundation model and does not remove online projection compute. Improved low-IPC results do not imply dominance at every IPC, and external pretraining is an additional information source.

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

2, 6

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 · 10 pages