Writing study: Online Reasoning Video Segmentation with Just-in-Time Digital Twins

ICCV overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICCV 2025 oral writing corpus

Yiqing Shen, Bohan Liu, Chenjia Li, Lalithkumar Seenivasan, Mathias Unberath · Official ICCV 2025 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: Architecture & representation (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

9 physical pages

Before references / appendix

≈ 8.05 page-equivalents

Main figures / tables

4 / 6

Attached appendix figures / tables

0 / 0

Attached appendix span

No boundary detected; not assumed zero

Before references / appendix: ≈ 89.44%References / intervening material: ≈ 10.56%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	≈ 1.16 pp.	12.89%
2. Related Works	2	≈ 0.52 pp.	5.78%
3. Methods	3	≈ 3.32 pp.	36.89%
4. Experiments	6	≈ 2.16 pp.	24%
5. Conclusion	8	≈ 0.26 pp.	2.89%
References	9	≈ 0.95 pp.	10.56%

All detected figure and table locations

Main text

• Table 1 · p. 2
• Figure 1 · p. 4
• Figure 2 · p. 5
• Figure 3 · p. 6
• Table 2 · p. 7
• Table 3 · p. 7
• Figure 4 · p. 7
• Table 4 · p. 8
• Table 5 · p. 8
• Table 6 · p. 8

Attached appendix

No recognized caption labels.

Argument and organization

1. 1 Introduction; 2 Related Works

   Motivates separating high-level query reasoning from specialist perception for online video segmentation.

2. 3 Methods

   Defines planning, scene-graph state, execution and a benchmark with categorized reasoning complexity.

3. 4 Experiments; 5 Conclusion

   Tests video/image reasoning, component removal and alternative planner/coder model configurations.

Problem, claim and contribution

The digital twin is an incrementally updated query-relevant scene graph, not a photorealistic or physically simulated replica.

Sections 3.1–3.3, physical pp.3–5 · Main paper

Argument and section order

Planning is separated from streaming execution and state maintenance, making the temporal information path inspectable.

Section 3 and Figure 1, pp.3–5 · Main paper

Figures, captions and visual explanation

The overview traces one flamingo query through model selection, graph nodes, intermediate perception outputs and the final mask, rather than presenting only a black-box agent.

Figure 1, p.4 · Main paper

Tables, comparisons and evidence

The benchmark reports category and difficulty cells, but 895 queries come from 200 videos and only 77 are temporal; image-method baselines are adapted frame by frame.

Section 3.4 and Figure 3, p.6; Table 2, p.7 · Main paper

Appendix and reproducibility

The main paper supplies the benchmark construction, default six-frame window and GPT model roles directly; no uninspected appendix is used to substantiate reproducibility or speed.

Section 3.4 and implementation details, p.6 · Main paper

Clarity and writing conventions

The LLM ablation explicitly trades higher performance from GPT-4o in both roles against stated overhead and chooses a mixed configuration.

Table 6 and Section 4.3, p.8 · Main paper

What to reuse—and what not to infer

Transferable lesson: For an agent system, show the planning/execution boundary and the maintained state, then evaluate both task complexity and component dependence.

Boundary: Online ordering is not measured real-time throughput: the main tables do not quantify latency, API cost or code-generation failure. The benchmark is imbalanced and queries share videos; pretrained specialist and proprietary LLM changes constrain reproducibility. Segmentation performance does not demonstrate embodied action safety or physically faithful digital twins.

Apply this lesson using the argument and evidence worksheet.

Inspection record and sources

Checked

2026-10-07

Reading status

Writing-focused manuscript scan + selected visuals inspected

Main text

Writing-oriented scan across the main manuscript, physical PDF pp.1–8, covering framing, section progression, method explanations, captions, comparison protocols and concluding scope; reference boundary checked. Selected full-page rasters were inspected. Not exhaustive line-by-line reading, proof verification or experimental reproduction.

Appendix / supplement

Main-paper appendix/supplement references and the reference boundary were checked. No separate supplementary link appears in this dated CVF inventory; this does not establish that no supplementary material exists elsewhere.

Visually inspected PDF pages

4, 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.

• Official CVF open-access ICCV 2025 accepted-paper PDF, retrieved 7 October 2026; CVF states identical to accepted version except watermark; not assumed identical to initial submission · 9 pages