Writing study: WorldLens: Full-Spectrum Evaluations of Driving World Models in Real World

CVPR overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← CVPR 2026 oral writing corpus

Ao Liang, Lingdong Kong, Tianyi Yan, Hongsi Liu, Yu Yang, Ziqi Huang, Wei Yin, Jialong Zuo, Yixuan Hu, Dekai Zhu, Dongyue Lu, Youquan Liu, Guangfeng Jiang, Linfeng Li, Xiangtai Li, Long Zhuo, Lai Xing Ng, Benoit R. Cottereau, Changxin Gao, Liang Pan, Wei Tsang Ooi, Ziwei Liu · 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: Datasets, benchmarks & measurement (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

15 physical pages

Before references / appendix

≈ 8.17 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: ≈ 54.47%References / intervening material: ≈ 45.53%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	≈ 0.81 pp.	5.4%
2. Related Work	2	≈ 0.67 pp.	4.47%
3. WorldLens : A Full-Spectrum Benchmark	3	≈ 2.25 pp.	15%
4. Human Annotation & Evaluation Agent	5	≈ 0.7 pp.	4.67%
5. Experiments	6	≈ 2.54 pp.	16.93%
6. Conclusion	8	≈ 0.18 pp.	1.2%
Acknowledgments	9	≈ 0.13 pp.	0.87%
References	9	≈ 6.83 pp.	45.53%

All detected figure and table locations

Main text

• Figure 1 · p. 1
• Figure 2 · p. 3
• Figure 3 · p. 4
• Figure 4 · p. 5
• Table 1 · p. 6
• Figure 5 · p. 6
• Figure 6 · p. 7
• Table 2 · p. 7
• Table 3 · p. 7
• Figure 7 · p. 7
• 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 62 physical pages; 29 detected figures and 27 tables. These may include overlapping main-text material.

Located appendix only: starts on physical p.1; ≈ 61.88 page-equivalents, 29 figure labels and 27 table labels. This interval is distinct from the whole file.

Detected heading	PDF page	Estimated span
Supplementary Material	1	≈ 0.97 pp.
A. Aspect 1: Generation	2	≈ 16 pp.
B. Aspect 2: Reconstruction	18	≈ 8 pp.
C. Aspect 3: Action-Following	26	≈ 8 pp.
D. Aspect 4: Downstream Task	34	≈ 8 pp.
E. Aspect 5: Human Preference	42	≈ 12 pp.
F. Evaluation Agent	54	≈ 5 pp.
G. Broader Impact & Limitations	59	≈ 0.51 pp.
H. Public Resource Used	59	≈ 1.49 pp.
References	61	≈ 1.91 pp.

• Table I · p. 2
• Figure I · p. 3
• Table II · p. 4
• Figure II · p. 5
• Table III · p. 6
• Figure III · p. 7
• Table IV · p. 8
• Figure IV · p. 9
• Table V · p. 10
• Figure V · p. 11
• Table VI · p. 12
• Figure VI · p. 13
• Table VII · p. 14
• Figure VII · p. 15
• Table VIII · p. 16
• Figure VIII · p. 17
• Table IX · p. 18
• Figure IX · p. 19
• Table X · p. 20
• Figure X · p. 21
• Table XI · p. 22
• Figure XI · p. 23
• Table XII · p. 24
• Figure XII · p. 25
• Table XIII · p. 26
• Figure XIII · p. 27
• Table XIV · p. 28
• Figure XIV · p. 29
• Table XV · p. 30
• Figure XV · p. 31
• Table XVI · p. 32
• Figure XVI · p. 33
• Table XVII · p. 34
• Figure XVII · p. 35
• Table XVIII · p. 36
• Figure XVIII · p. 37
• Table XIX · p. 38
• Figure XIX · p. 39
• Table XX · p. 40
• Figure XX · p. 41
• Table XXI · p. 42
• Figure XXI · p. 43
• Table XXII · p. 44
• Figure XXII · p. 45
• Table XXIII · p. 46
• Figure XXIII · p. 47
• Table XXIV · p. 48
• Figure XXIV · p. 49
• Table XXV · p. 50
• Figure XXV · p. 51
• Table XXVI · p. 52
• Figure XXVI · p. 53
• Figure XXVII · p. 54
• Table XXVII · p. 55
• Figure XXVIII · p. 57
• Figure XXIX · p. 58

Read-only audit 2026-09-26: restored actual supplementary section headings from cached PDF layout, excluding table-of-contents entries and retaining physical PDF page coordinates. This measures reading-order spans, not occupied text area. Roman-numeral captions are present; Figure I on p3 was visually inspected. Other Roman labels use source text coordinates, not independent visual checks.

Argument and organization

1. 3. WorldLens Benchmark

   Defines five evaluation aspects and their measurable dimensions.

2. 4–5. Annotation, Agent, and Experiments

   Explains human scoring and agent training before per-aspect comparisons and insights.

Problem, claim and contribution

The opening distinguishes looking realistic from behaving plausibly, and explicitly reports that no model dominates every axis.

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

Argument and section order

Stable aspect/dimension codes carry the taxonomy from definitions into tables and supplementary sections.

§3 pp. 3–5; Tables 1–3 · Main paper · Additional source 2

Figures, captions and visual explanation

The annotation UI shows video, semantics, depth, and 3D boxes together; a downstream grid aligns generated-world defects with perception outputs.

Figure 3 p. 4; Figure 6 p. 7 · Main paper · Additional source 2

Tables, comparisons and evidence

Action-following and downstream-task tables remain separate; the prose highlights poor closed-loop behavior despite more favorable open-loop scores.

Tables 2–3 and §5.1 p. 7 · Main paper · Additional source 2

Appendix and reproducibility

The long supplement is navigable because each dimension repeats definition, formulation, implementation, examples, and analysis rather than collecting an undifferentiated gallery.

Supplement §§A–E · Main paper · Additional source 2

Clarity and writing conventions

Human-preference summary plots show distribution summaries, and the limitations identify annotator bias and evaluator-model dependence.

Figure 7 p. 7; supplement §G.3 · Main paper · Additional source 2

What to reuse—and what not to infer

Transferable lesson: For a broad benchmark, use a stable taxonomy and a repeatable per-metric evidence template; preserve disagreements between dimensions.

Boundary: Human-perceived behavioral safety and an agent's agreement with those ratings are not certified driving safety.

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 the 62-page supplement at writing-structure level: aspect/dimension organization, repeated definition–formulation–implementation–example–analysis units, human-score rubrics, agent protocol, limitations and resources. Not an independent audit of every metric computation.

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 · 15 pages
• Official CVPR 2026 separate supplementary PDF; retrieved 2026-09-26 · 62 pages