Writing study: Delving into Latent Spectral Biasing of Video VAEs for Superior Diffusability
← ECCV 2026 oral writing corpus
Shizhan Liu, Xinran Deng, Zhuoyi Yang, Jiayan Teng, Xiaotao Gu, Jie Tang · Official ECCV 2026 oral record
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
- 18 physical pages
- Before references / appendix
- ≈ 15.15 page-equivalents
- Main figures / tables
- 6 / 6
- 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 pp. | 11.11% |
| 2 Related Work | 3 | ≈ 0.87 pp. | 4.83% |
| 3 Channel Eigenspectrum Shaping | 4 | ≈ 3.31 pp. | 18.39% |
| 4 Spatio-Temporal Frequency Spectrum Shaping | 7 | ≈ 2.48 pp. | 13.78% |
| 5 Experiments | 10 | ≈ 5.49 pp. | 30.5% |
| 6 Conclusion | 15 | ≈ 0.43 pp. | 2.39% |
| References | 16 | ≈ 2.85 pp. | 15.83% |
All detected figure and table locations
Main text
Attached appendix
No recognized caption labels.
Argument and organization
- 1 Introduction; 2 Related Work
Separates VAE reconstruction quality from downstream diffusability and motivates two distinct spectral properties rather than treating latent quality as one score.
- 3 Channel Eigenspectrum Shaping; 4 Spatio-Temporal Frequency Spectrum Shaping
Each spectral branch follows diagnostic observation, explanatory analysis and a tractable training intervention before the joint experiment section.
- 5 Experiments; 6 Conclusion
Incremental component tables distinguish regularizers and decoder fine-tuning, while separate reconstruction tables keep generative benefits from hiding fidelity losses.
Problem, claim and contribution
Separates VAE reconstruction quality from downstream diffusability and motivates two distinct spectral properties rather than treating latent quality as one score.
Abstract and Section 1, pp.1–2 · Main paper
Argument and section order
Each spectral branch follows diagnostic observation, explanatory analysis and a tractable training intervention before the joint experiment section.
Sections 3–4, pp.4–9 · Main paper
Figures, captions and visual explanation
The masked-reconstruction schematic is paired with explained-variance and eigenspectrum plots; the common placement connects a training operation to its intended statistical effect.
Figure 3, p.7 · Main paper
Tables, comparisons and evidence
Incremental component tables distinguish regularizers and decoder fine-tuning, while separate reconstruction tables keep generative benefits from hiding fidelity losses.
Tables 2–3, p.13; Tables 4–6, p.15 · Main paper
Appendix and reproducibility
The theorem's proof and scope are explicitly sent to Supplement Section 1; further pointers cover correlation derivations, excess bias and overhead. Those separate sections were not inspected.
Theorem 1, p.6; Sections 4 and 6, pp.8–9,15 · Main paper
Clarity and writing conventions
Names eigenspectrum and spatio-temporal spectrum consistently and marks ablation training budgets separately from main-model training.
Sections 3–4 and 5.3, pp.4–9,13 · Main paper
What to reuse—and what not to infer
Transferable lesson: For a regularizer, show the diagnosed distribution, the operation that changes it, and both downstream quality and representation/reconstruction cost.
Boundary: A linear analysis of covariance is not a proof of every nonlinear diffusion-model outcome; convergence and reconstruction claims have distinct training settings.
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
- 7, 13
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.