Writing study: VGGT-Ω
← CVPR 2026 oral writing corpus
Jianyuan Wang, Minghao Chen, Shangzhan Zhang, Nikita Karaev, Johannes Schönberger, Patrick Labatut, Piotr Bojanowski, David Novotny, Andrea Vedaldi, Christian Rupprecht · Official CVPR 2026 oral record
Figure, table and section ledger
Editorial contribution type: Architecture & representation (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
- 14 physical pages
- Before references / appendix
- ≈ 8.05 page-equivalents
- Main figures / tables
- 4 / 3
- Attached appendix figures / tables
- 0 / 0
- Attached appendix span
- Separate file recorded below; not in this PDF total
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.
All detected figure and table locations
Main text
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 3 physical pages; 0 detected figures and 0 tables. These may include overlapping main-text material.
Located appendix only: starts on physical p.1; ≈ 2.94 page-equivalents, 0 figure labels and 0 table labels. This interval is distinct from the whole file.
| Detected heading | PDF page | Estimated span |
|---|---|---|
| Supplementary Material | 1 | ≈ 0.07 pp. |
| A. Additional Details | 1 | ≈ 1.91 pp. |
| B. Discussions | 3 | ≈ 0.26 pp. |
| C. Limitations | 3 | ≈ 0.46 pp. |
| References | 3 | ≈ 0.24 pp. |
Argument and organization
- Scaling visual geometry §3
Explain architecture and losses, then dynamic reconstruction, self-supervision and data annotation.
- Experiments and applications §§4–5
Benchmark static/dynamic camera/depth tasks, isolate scale/design choices and test frozen scene tokens in a VLA.
Problem, claim and contribution
The opening asks whether model and data scaling benefit feed-forward geometry, then identifies efficiency and annotation quality as enabling obstacles rather than claiming size alone is the method.
Abstract and Figure 1, PDF pp. 1–2 · Main paper · Additional source 2
Argument and section order
The paper separates output heads from training-only losses and gives the annotation pipeline a method subsection, treating data construction as part of the scientific mechanism.
§§3.1–3.2, PDF pp. 3–4; §3.5, PDF p. 6 · Main paper · Additional source 2
Figures, captions and visual explanation
The architecture labels point/matching losses as training only and explicitly shows the alternative scene-token attention path, so removed inference heads cannot be mistaken for missing supervision.
Figure 2, PDF p. 3 · Main paper · Additional source 2
Tables, comparisons and evidence
Camera and depth tables separate static and dynamic datasets; the scaling analysis states approximately matched token counts, and the annotation-quality check discloses filtered sequences and exclusion of dynamic regions.
Tables 1–2, PDF p. 7; §4.3, PDF p. 8 · Main paper · Additional source 2
Appendix and reproducibility
The concise supplement documents normalization, loss weights, augmentation/pairing and filtering prompts, then discusses simplifications and depth-label ambiguities across datasets.
Supplement §§A–C, PDF pp. 1–3 · Main paper · Additional source 2
Clarity and writing conventions
The scaling discussion says the curves suggest power laws rather than presenting an established law; the supplement also acknowledges accuracy deliberately traded for architectural simplicity.
§4.3, PDF p. 8; Supplement §B, PDF p. 3 · Main paper · Additional source 2
What to reuse—and what not to infer
Transferable lesson: A scaling paper should disclose the controlled compute quantity, explain the pipeline that enables scale and distinguish empirical scaling trends from universal laws.
Boundary: Matched tokens are not identical FLOPs across model sizes; filtered static annotation checks do not independently validate every dynamic scene, and internal data sources constrain exact replication.
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 PDF pp. 1–14: framing, argument sequence, section organization, figure/table captions, evaluation interpretation and limitations; reference sections identified. This is a writing-focused scan, not an exhaustive technical proof or reproduction audit. Main PDF pages 3 and 6 were rendered and visually inspected.
- Appendix / supplement
- Writing-focused scan of the separate supplementary PDF, pp. 1–3, including its organization, captions, protocol/evidence detail and limitation material where present; proofs and code listings were inspected for exposition, not independently verified or executed. Supplement §§A–C, PDF pp. 1–3. Linked video/audio demonstrations and non-PDF archive contents were not assessed.
- Visually inspected PDF pages
- 3, 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.