Writing study: RefAV: Towards Planning-Centric Scenario Mining
← CVPR 2026 oral writing corpus
Cainan Davidson, Deva Ramanan, Neehar Peri · Official CVPR 2026 oral record
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
- 12 physical pages
- Before references / appendix
- ≈ 8.04 page-equivalents
- Main figures / tables
- 4 / 4
- 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.
| Top-level heading | Starts on PDF page | Estimated span | Share of entire PDF |
|---|---|---|---|
| 1. Introduction | 1 | ≈ 1.58 pp. | 13.17% |
| 2. Related Works | 2 | ≈ 0.78 pp. | 6.5% |
| 3. RefAV: Scenario Mining with Natural Lan- | 3 | ≈ 3.1 pp. | 25.83% |
| 4. Experiments | 6 | ≈ 2.05 pp. | 17.08% |
| 5. Conclusion | 8 | ≈ 0.09 pp. | 0.75% |
| Acknowledgments | 8 | ≈ 0.13 pp. | 1.08% |
| References | 9 | ≈ 3.96 pp. | 33% |
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 16 physical pages; 4 detected figures and 3 tables. These may include overlapping main-text material.
Located appendix only: starts on physical p.1; ≈ 15.96 page-equivalents, 4 figure labels and 3 table labels. This interval is distinct from the whole file.
Argument and organization
- RefAV §3.1
Explain the source data, planning scenarios, compositional programs and human verification.
- Baselines and experiments §§3.2–4
Define progressively richer mining strategies and separate perception, temporal localization and retrieval evaluation.
Problem, claim and contribution
Scenario mining is distinguished from ordinary referring-object tracking because a described event may be absent and must be localized in both 3D space and time.
Introduction, PDF p. 1; Figure 1, PDF p. 2 · Main paper · Additional source 2
Argument and section order
Dataset creation precedes the zero-shot baseline suite; the method explanation distinguishes human-corrected annotation programs from unverified inference-time generated programs.
§§3.1–3.2, PDF pp. 3–6 · Main paper · Additional source 2
Figures, captions and visual explanation
The dual-path diagram separates offline 3D perception from prompt-to-program generation, then joins them at code execution; referred, related and other objects retain consistent roles across diagrams.
Figure 4, PDF p. 6; Figures 1–3, PDF pp. 2, 4–5 · Main paper · Additional source 2
Tables, comparisons and evidence
The benchmark table retains a simple referred-class filter and includes oracle ground-truth tracks beside predicted trackers, exposing both retrieval difficulty and upstream perception error.
Table 2 and §4.1, PDF p. 7 · Main paper · Additional source 2
Appendix and reproducibility
The supplement includes baseline implementation, token statistics, annotation interface, sampling-rate analysis, positive/negative scenario distributions, generated examples and an API listing.
Supplement §§A–J, PDF pp. 1–16, Figures 5–8 · Main paper · Additional source 2
Clarity and writing conventions
The metrics paragraph explains how HOTA-Temporal penalizes timing mistakes differently from HOTA-Track, and program validity is reported separately from mining accuracy.
§4, PDF p. 6; Table 3, PDF p. 8 · Main paper · Additional source 2
What to reuse—and what not to infer
Transferable lesson: For compositional retrieval, make negative examples and temporal boundaries part of the task definition, and separate upstream perception quality from downstream reasoning quality.
Boundary: Oracle-track results are upper-bound conditions, not raw-sensor deployment performance; executing valid code does not imply it denotes the intended scenario.
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. 1–12: 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. 1–16, 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–J, PDF pp. 1–16, Figures 5–8. 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.