Writing study: Understanding Geometric Representations in Self-Supervised Vision Transformers via Subspace Intervention
← ECCV 2026 oral writing corpus
Weichen Zhou, Yawen Zou, Chunzhi Gu, Ran Dong, Haoran Xie, Chao Zhang · Official ECCV 2026 oral record
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
- 18 physical pages
- Before references / appendix
- ≈ 14.59 page-equivalents
- Main figures / tables
- 8 / 2
- Attached appendix figures / tables
- 0 / 0
- Attached appendix span
- No boundary detected; not assumed zero
Extraction notes: No attached appendix boundary detected. Zero means no recognized caption, not independently verified absence.
Source-bound correction: All top-level main headings and the reference boundary checked against source PDF text; source-bound positions use physical single-column y/page-height, not a two-column reading-order surrogate. Running page-number/author titles, affiliations, numbered contribution bullets, footnotes and bibliography continuations are excluded. This linked manuscript contains no appended supplementary section; separate supplement links were not scanned.
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 | ≈ 2.06 pp. | 11.44% |
| 2 The Diagnostic Framework | 3 | ≈ 2.59 pp. | 14.39% |
| 3 Experimental Design and Objectives | 6 | ≈ 1.4 pp. | 7.78% |
| 4 Results and Analysis | 7 | ≈ 6.41 pp. | 35.61% |
| 5 Related Work | 14 | ≈ 1 pp. | 5.56% |
| 6 Conclusion | 15 | ≈ 0.36 pp. | 2% |
| Acknowledgements | 15 | ≈ 0.08 pp. | 0.44% |
| References | 15 | ≈ 3.41 pp. | 18.94% |
All detected figure and table locations
Main text
Attached appendix
No recognized caption labels.
Argument and organization
- 1 Introduction; 2 The Diagnostic Framework
Distinguishes absent information from information a particular probe cannot read, then turns that ambiguity into controlled subspace interventions.
- 3 Experimental Design and Objectives; 4 Results and Analysis
Defines probe hierarchy and random/residual controls before an explicit analytical roadmap; the result subsections follow the declared questions.
- 5 Related Work; 6 Conclusion
Random and orthogonal-residual controls test whether retaining dimensions alone explains recovery; separate layers, tasks and robustness analyses constrain the interpretation.
Problem, claim and contribution
Distinguishes absent information from information a particular probe cannot read, then turns that ambiguity into controlled subspace interventions.
Sections 1–2, pp.1–5 · Main paper
Argument and section order
Defines probe hierarchy and random/residual controls before an explicit analytical roadmap; the result subsections follow the declared questions.
Sections 2–3, pp.3–7 · Main paper
Figures, captions and visual explanation
The framework keeps the backbone and readout fixed while varying projected subspaces. Paired rank curves separate absolute task accuracy from normalized recovery efficiency.
Figure 1, p.3; Figure 2, p.8 · Main paper
Tables, comparisons and evidence
Random and orthogonal-residual controls test whether retaining dimensions alone explains recovery; separate layers, tasks and robustness analyses constrain the interpretation.
Section 2.3, p.5; Sections 4.2–4.5, pp.8–13 · Main paper
Appendix and reproducibility
Hyperparameters, detailed probe architectures and stability metrics are pointed to in supplemental material; these are not counted as separately read sources.
Sections 3.2 and 4.3–4.5, pp.7,11,13 · Main paper
Clarity and writing conventions
The roadmap explains what each analysis can answer; limitations distinguish explicitly linear structure from nonlinear entanglement and broader domain validation.
Section 3.3, p.7; Limitations, p.15 · Main paper
What to reuse—and what not to infer
Transferable lesson: A representation analysis benefits from a question-led roadmap, frozen controls and separate absolute and normalized plots; label what the probe can and cannot establish.
Boundary: Low-rank linear recoverability does not prove that all geometric information lives only in that subspace, nor that the same result holds across untested domains.
Apply this lesson using the argument and evidence worksheet.
Inspection record and sources
- Checked
- 2026-10-02
- Reading status
- Writing-focused manuscript scan + selected visuals inspected
- Main text
- Writing-oriented scan across the manuscript main argument, physical PDF pp.1–15: framing and section roadmaps, method/argument explanations, figure and table captions, comparison protocols, conclusions and limitations. Selected full-page rasters were inspected. This is not exhaustive line-by-line technical reading, proof verification or experimental reproduction.
- Appendix / supplement
- Main-manuscript pointers to appendix/supplement and the reference boundary were checked. Separate supplementary PDFs were not retrieved or inspected; their contents are not included in inspected-page counts.
- Visually inspected PDF pages
- 3, 8
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.