Writing study: Wan-Weaver: Interleaved Multi-modal Generation via Decoupled Training

CVPR overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← CVPR 2026 oral writing corpus

Jinbo Xing, Zeyinzi Jiang, Yuxiang Tuo, Chaojie Mao, Xiaotang Gai, Xi Chen, Jingfeng Zhang, Yulin Pan, Zhen Han, Jie Xiao, Keyu Yan, Chenwei Xie, Chongyang Zhong, Kai Zhu, Tong Shen, Lianghua Huang, Yu Liu, Yujiu Yang · 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: 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

12 physical pages

Before references / appendix

≈ 8.05 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: ≈ 67.08%References / intervening material: ≈ 32.92%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	2	≈ 0.73 pp.	6.08%
2. Related Work	2	≈ 0.92 pp.	7.67%
3. Method	3	≈ 2.2 pp.	18.33%
4. Experiment	5	≈ 2.96 pp.	24.67%
5. Conclusion	8	≈ 0.19 pp.	1.58%
References	9	≈ 3.95 pp.	32.92%

All detected figure and table locations

Main text

• Figure 1 · p. 1
• Figure 2 · p. 3
• Figure 3 · p. 4
• Table 1 · p. 6
• Figure 4 · p. 6
• Table 2 · p. 6
• Figure 5 · p. 7
• Table 3 · p. 7
• Figure 6 · p. 8
• Figure 7 · p. 8
• Figure 8 · 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 18 physical pages; 7 detected figures and 8 tables. These may include overlapping main-text material.

Located appendix only: starts on physical p.4; ≈ 14.23 page-equivalents, 7 figure labels and 5 table labels. This interval is distinct from the whole file.

Detected heading	PDF page	Estimated span
10. Importantly, unlike prior benchmarks that rely on GPT-	4	≈ 13.28 pp.
References	18	≈ 0.95 pp.

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

Argument and organization

1. 3. Method

   Separates inference architecture, decoupled optimization, and data curation.

2. 4. Experiment

   Defines WeaverBench before interleaved/single-modality results and training/data ablations.

Problem, claim and contribution

The gap is producing interleaved outputs with long-range consistency, not merely accepting multimodal inputs.

Abstract; §1 · Main paper · Additional source 2

Argument and section order

Inference and training have separate diagrams, avoiding a common confusion in systems with independently optimized components.

Figures 2–3 pp. 3–4 · Main paper · Additional source 2

Figures, captions and visual explanation

The training diagram marks frozen and trainable modules and names their objectives; the qualitative comparison preserves alternating text and images so document-level consistency is visible.

Figure 3 p. 4; Figure 5 p. 7 · Main paper · Additional source 2

Tables, comparisons and evidence

The analysis retains a commercial comparator's overall lead in one setting and distinguishes interleaved capabilities from separate understanding/generation/editing scores.

§4.3 and Table 3 p. 7 · Main paper · Additional source 2

Appendix and reproducibility

Benchmark scoring prompts and tie conventions accompany extensive failure examples, including collapsed layouts and incorrect geometric reasoning.

Supplement §§B,F; Table 4 · Main paper · Additional source 2

Clarity and writing conventions

The ablation section identifies its smaller 7B variant as a compute-driven choice, rather than implying identical-scale evidence.

§4.5 p. 7 · Main paper · Additional source 2

What to reuse—and what not to infer

Transferable lesson: For unified generation, evaluate the composition of modalities explicitly, not just the quality of each modality separately.

Boundary: Automated judge preferences and selected storyboards are proxies; preserve rubric, ties, model variant, and documented failure modes.

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 18 supplementary pages spanning benchmark categories/rubrics, judge prompts and win-rate conventions, extra task results, failures, and extended generation examples.

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 · 12 pages
• Official CVPR 2026 separate supplementary PDF; retrieved 2026-09-26 · 18 pages