Writing study: XR-1: Towards Versatile Vision-Language-Action Models via Learning Unified Vision-Motion Representations

ICML overviewCorpus atlasBy contribution typeVisual galleryRules & templateWriting patternsPaper studies

← ICML 2026 oral writing corpus

Shichao Fan, Kun Wu, Zhengping Che, Xinhua Wang, Di Wu, Fei Liao, Ning Liu, Yixue Zhang, Zhen Zhao, Zhiyuan Xu, Meng Li, Qingjie Liu, Shanghang Zhang, Min Wan, Jian Tang · Official ICML 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: 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

45 physical pages

Before references / appendix

≈ 9.31 page-equivalents

Main figures / tables

7 / 4

Attached appendix figures / tables

11 / 15

Attached appendix span

≈ 29.91 page-equivalents

Before references / appendix: ≈ 20.69%References / intervening material: ≈ 12.84%Attached appendix: ≈ 66.47%Estimated layout envelopes, not text/artwork area. Unresolved appendix boundaries remain within the undivided post-body material.

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	≈ 0.92 pp.	2.04%
2. Related Work	3	≈ 1.11 pp.	2.47%
3. Methodology	4	≈ 2.09 pp.	4.64%
4. Experiments	6	≈ 3.49 pp.	7.76%
5. Conclusion	9	≈ 0.24 pp.	0.53%
Acknowledgments	10	≈ 0.21 pp.	0.47%
Impact Statement	10	≈ 0.05 pp.	0.11%
References	10	≈ 5.78 pp.	12.84%
A. Open-Source Resources	16	≈ 0.11 pp.	0.24%
B. Implementation Details	16	≈ 2.6 pp.	5.78%
C. Dataset Curation	18	≈ 2.29 pp.	5.09%
D. Additional Real-World Experiments	21	≈ 2.3 pp.	5.11%
E. Additional Ablation Study	23	≈ 1.12 pp.	2.49%
F. Additional Generalization Analysis	24	≈ 0.98 pp.	2.18%
G. Simulation Results on CALVIN	25	≈ 0.22 pp.	0.49%
H. Visualization of Unified Vision-Motion Codes	25	≈ 3.91 pp.	8.69%
I. Failure Case Analysis on Baseline Methods	29	≈ 0.93 pp.	2.07%
J. Failure Case Analysis on XR-1	30	≈ 0.11 pp.	0.24%
K. Representative Tasks	30	≈ 15.35 pp.	34.11%

All detected figure and table locations

Main text

• Figure 1 · p. 1
• Figure 2 · p. 4
• Table 1 · p. 5
• Figure 3 · p. 6
• Figure 4 · p. 6
• Table 2 · p. 7
• Table 3 · p. 7
• Table 4 · p. 8
• Figure 5 · p. 8
• Figure 6 · p. 9
• Figure 7 · p. 9

Attached appendix

• Table 5 · p. 17
• Table 6 · p. 17
• Table 7 · p. 18
• Figure 8 · p. 19
• Table 8 · p. 19
• Table 9 · p. 21
• Table 10 · p. 21
• Table 11 · p. 22
• Table 12 · p. 22
• Table 13 · p. 23
• Table 14 · p. 23
• Table 15 · p. 24
• Table 16 · p. 24
• Figure 9 · p. 24
• Figure 10 · p. 25
• Table 17 · p. 26
• Figure 11 · p. 26
• Figure 12 · p. 27
• Figure 13 · p. 28
• Table 18 · p. 29
• Figure 14 · p. 29
• Figure 15 · p. 30
• Figure 16 · p. 31
• Figure 17 · p. 31
• Figure 18 · p. 32
• Table 19 · p. 33

Argument and organization

1. 1 Introduction; 2 Related Work

   Motivates a joint visual-dynamics and motor representation rather than either unimodal latent.

2. 3 Methodology

   Organizes the representation learner, generalist pretraining and deployment refinement as three stages.

3. 4 Experiments; 5 Conclusion

   Separates robot comparisons, component ablations, shifted scenarios, out-of-box use and few-shot adaptation.

Problem, claim and contribution

Two stated bottlenecks map to a shared code space, but the claim of causal semantics is stronger than the reconstruction and alignment objectives alone establish.

Introduction and Related Work, pp.2-3; objectives, p.5 · Main paper

Argument and section order

Each training stage specifies its data and loss, including a reduced visual-only objective for human videos without action labels.

Sections 3.2-3.4, pp.4-5 · Main paper

Figures, captions and visual explanation

The overview separates the three training stages and their data sources; result bars distinguish out-of-box and adapted variants and isolate averages from individual tasks.

Figure 2, p.4; Figures 6-7, p.9 · Main paper

Tables, comparisons and evidence

Out-of-box means omitting Stage 3, not an unseen pretraining embodiment; novel-embodiment adaptation uses demonstrations, and the caption discloses joint versus per-task baseline training.

Section 4.4, p.8; Figure 7 caption, p.9 · Main paper

Appendix and reproducibility

The supplement exposes initialization and implementation choices, codebook and modality ablations, and specific precision/coordination failures.

Appendix B, p.17; Appendix E, p.23; Appendix J, p.30 · Main paper

Clarity and writing conventions

Phrases such as monotonic scaling law and embodiment-agnostic semantics should remain tied to the finite tested data fractions and transfer protocols.

Sections 4.3-4.5, pp.8-9 · Main paper

What to reuse—and what not to infer

Transferable lesson: Label representation pretraining, policy pretraining and task adaptation separately so transfer claims reveal exactly which data were available.

Boundary: Unseen during pretraining is not zero-shot deployment; sampled rollout scores and visualization clusters do not establish universal transfer or causal latent semantics.

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 of the main argument through physical p.9, with focused physical-page text inspection on 1, 3, 4, 5, 6, 7, 8, 9. Framing, section progression, mechanism/proof explanation, displays, comparison conditions and conclusions were inspected. This is not exhaustive technical reading, proof verification or reproduction.

Appendix / supplement

Appendix organization and main-text pointers were checked; focused text inspection of physical pages 17, 23, 30. Remaining appendix proofs and details were not exhaustively read. No separately linked supplement inspected.

Visually inspected PDF pages

4, 9

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.

• Published proceedings PDF, PMLR volume 306, retrieved 2026-10-07 · 45 pages