Skip to content
Computational Pathology Reading Roadmap

Writing study: The Stability of Online Algorithms in Performative Prediction

← NeurIPS 2026 oral writing corpus

Gabriele Farina, Juan C Perdomo · Official NeurIPS 2026 oral record

Figure, table and section ledger

Editorial contribution type: Theory & formal analysis (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
18 physical pages
Before references / appendix
≈ 15.09 page-equivalents
Main figures / tables
2 / 1
Attached appendix figures / tables
0 / 0
Attached appendix span
No boundary detected; not assumed zero
Before references / appendix: ≈ 83.83%References / intervening material: ≈ 16.17%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.

Source-bound correction: Top-level heading and reference/appendix boundaries checked against source PDF text. Bibliography titles/author initials, numbered contribution/property bullets, diagram steps, equations and TOC entries are not substantive headings. Section spans are page-coordinate estimates, not ink area or prose-only length. This is an author manuscript mirror, not a verified accepted-version template. Source layout uses physical single-column y/page-height, including equation-heavy pages.

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 headingStarts on PDF pageEstimated spanShare of entire PDF
1 Introduction2≈ 4 pp.22.22%
2 Preliminaries and Technical Background6≈ 2.3 pp.12.78%
3 Performative Stability of Online Algorithms8≈ 5.7 pp.31.67%
4 Analyzing the Multiplayer Case14≈ 1.5 pp.8.33%
5 Discussion and Future Work15≈ 0.5 pp.2.78%
References16≈ 2.91 pp.16.17%
All detected figure and table locations

Main text

Attached appendix

    No recognized caption labels.

    Argument and organization

    1. 1 Introduction

      Defines feedback and motivates stability of a model mixture.

    2. 2 Preliminaries and Technical Background

      Separates stability from optimality and recalls online-learning background.

    3. 3 Performative Stability of Online Algorithms

      Proves a reduction and instantiates it for several algorithms.

    4. 4 Analyzing the Multiplayer Case

      Extends the martingale/regret argument to multiple learners.

    5. 5 Discussion and Future Work

      States remaining optimality and stateful-feedback questions.

    Problem, claim and contribution

    The introduction identifies the prediction-to-data feedback loop, then explains why a mixture of deployed models changes the stability question. The informal result is presented with that changed solution concept rather than merely advertising fewer assumptions.

    PDF pp.2–4, Introduction and informal Theorem 1 · Main paper

    Argument and section order

    Definitions and a visual informal claim precede prior-work comparisons and technical background. The main reduction is proved, instantiated for several algorithmic regimes, extended to multiple players, and then limited against risk optimality and stateful feedback.

    PDF pp.3–15, Sections 1–5 · Main paper

    Figures, captions and visual explanation

    Figure 1 combines a feedback-loop diagram with a boxed mathematical meaning of stability. Table 1 puts stable model and stable mixture on different rows and changes assumption regimes across columns; the layout makes the changed target explicit.

    PDF p.3, Figure 1; p.5, Table 1 · Main paper

    Tables, comparisons and evidence

    Theorem 3 states bounded-loss and regret requirements and its proof separates a martingale term from a regret term. The implications reuse that reduction for concrete settings; the discussion says stability need not minimize performative risk and leaves stateful environments open.

    PDF pp.8–13, Theorem 3 and implications; p.15, Discussion · Main paper

    Appendix and reproducibility

    This 18-page author manuscript has no separate appendix section: the main reduction proof, corollaries, and multiplayer argument remain in Sections 3–4, followed by references on pp.16–18. There is therefore no appendix content to count or invent for this source.

    PDF pp.8–18, Sections 3–5 and References · Main paper

    Clarity and writing conventions

    The comparison table defines its compact sensitivity ratio and qualifies rate comparisons up to logarithmic factors. The conclusion distinguishes stable from optimal and stateless from stateful instead of treating feedback as a single solved problem.

    PDF p.5, Table 1 caption; p.15, Discussion · Main paper

    What to reuse—and what not to infer

    Transferable lesson: When weakening assumptions by changing the solution concept, put both changes in the same comparison table and explain the resulting limits in the conclusion.

    Boundary: A stable mixture is not a single stable model, nor necessarily a performative-risk optimum. The public 18-page version is not evidence of an allowed NeurIPS submission length, and is not confirmed identical to accepted OpenReview.

    Apply this lesson using the argument and evidence worksheet.

    Inspection record and sources

    Checked
    2026-10-02
    Reading status
    Writing-focused manuscript scan + selected visuals inspected
    Main text
    Writing-oriented source scan of framing, argument roadmap, selected substantive passages, method/theorem statements, main figure/table captions, evidence and limits across PDF pp.1–15. Two selected main-body pages were actually inspected as rendered images. This is not exhaustive line-by-line technical verification or experiment reproduction. Title and all author names compared against official program metadata and manuscript first page; matching authors, with capitalization, middle initials or expanded given names where applicable.
    Appendix / supplement
    Source ending checked: Sections 3–4 contain proofs and pp.16–18 are references; no separate appendix was present in this manuscript.
    Visually inspected PDF pages
    3, 5

    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.