Writing study: FILTR: Extracting Topological Features from Pretrained 3D Models

CVPR overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← CVPR 2026 oral writing corpus

Louis Martinez, Maks Ovsjanikov · 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: Empirical analysis, diagnosis & auditing (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

≈ 8.08 page-equivalents

Main figures / tables

8 / 3

Attached appendix figures / tables

0 / 0

Attached appendix span

Separate file recorded below; not in this PDF total

Before references / appendix: ≈ 73.45%References / intervening material: ≈ 26.55%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	≈ 0.75 pp.	6.82%
2. Related works	2	≈ 0.64 pp.	5.82%
3. Do 3D encoders understand topology?	3	≈ 2.6 pp.	23.64%
4. FILTR	5	≈ 2.89 pp.	26.27%
5. Conclusion, limitations and future work	8	≈ 0.32 pp.	2.91%
References	9	≈ 2.92 pp.	26.55%

All detected figure and table locations

Main text

• Figure 1 · p. 1
• Figure 2 · p. 3
• Figure 3 · p. 4
• Table 1 · p. 4
• Figure 4 · p. 4
• Figure 5 · p. 5
• Figure 6 · p. 5
• Figure 7 · p. 6
• Table 2 · p. 6
• Figure 8 · p. 7
• Table 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 7 physical pages; 10 detected figures and 9 tables. These may include overlapping main-text material.

Located appendix only: starts on physical p.1; ≈ 6.94 page-equivalents, 10 figure labels and 9 table labels. This interval is distinct from the whole file.

Detected heading	PDF page	Estimated span
Supplementary Material	1	≈ 0.1 pp.
6. Implementation details	1	≈ 2.48 pp.
7. 3D vs. latent prediction pretraining	3	≈ 1.01 pp.
8. Experiments	4	≈ 3.35 pp.

• Figure 9 · p. 1
• Table 4 · p. 1
• Figure 10 · p. 2
• Table 5 · p. 3
• Figure 11 · p. 4
• Figure 12 · p. 4
• Figure 13 · p. 4
• Figure 14 · p. 4
• Table 6 · p. 5
• Figure 15 · p. 5
• Table 8 · p. 5
• Table 9 · p. 5
• Table 7 · p. 5
• Table 10 · p. 6
• Table 11 · p. 6
• Table 12 · p. 6
• Figure 16 · p. 7
• Figure 18 · p. 7
• Figure 17 · p. 7

Argument and organization

1. 3. Do 3D encoders understand topology?

   Constructs a controlled dataset and probes both labels and feature alignment.

2. 4. FILTR

   Turns the observed representation question into a set-prediction method and tests transfer.

Problem, claim and contribution

The introduction poses a testable question about topological information in pretrained features before presenting a decoder as the answer.

§1; Figure 1 · Main paper · Additional source 2

Argument and section order

Dataset label distributions and frozen-layer probes precede method design, so the intervention follows evidence rather than just a proposed architecture.

§3; Figures 2–6 · Main paper · Additional source 2

Figures, captions and visual explanation

The probing diagram marks frozen blocks and per-layer probes; the later variant diagram explicitly contrasts last-layer, combined-layer, and trainable encoder inputs.

Figure 4 p. 4; Figure 8 p. 7 · Main paper · Additional source 2

Tables, comparisons and evidence

The text acknowledges weak direct topology probes, then distinguishes them from successful persistence-diagram reconstruction using a learned decoder.

Table 1 p. 4; §4.5 p. 7 · Main paper · Additional source 2

Appendix and reproducibility

Controlled feature mismatch checks CKA interpretation, while per-category results and failure diagrams expose difficulty under domain shift.

Supplement §8.3 and Figures 14,17 · Main paper · Additional source 2

Clarity and writing conventions

The conclusion limits transfer to domains with suitable pretrained encoders rather than claiming topology understanding from successful decoding alone.

§5 p. 8 · Main paper · Additional source 2

What to reuse—and what not to infer

Transferable lesson: An analysis paper can build a persuasive question→controlled probe→informed method sequence while retaining negative findings.

Boundary: Probe accuracy, representational similarity, and recoverability through a trained decoder are distinct claims, not interchangeable evidence.

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 seven supplementary pages covering dataset construction algorithms, model/training settings, per-category probing, CKA mismatch controls, costs, decoder/loss ablations, and diagram failure cases.

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 · 7 pages