AAAI 2026 — corpus atlas

16 of 1058 oral records have a writing-focused manuscript study. Official program snapshot: 2026-10-25. Selected rendered pages were inspected for each study; this is not a proof audit or reproduction of experiments.

Teaching selection: 16 of 16 studies inspected. Choose an example by contribution style. Completing this selection does not mean the full oral inventory has been read.

Selected deep collection: the complete official main-track oral inventory is present, but only the counted studies have been read for writing analysis and visually inspected. Other records remain uninspected, not automatically scheduled or unavailable. This is not an all-paper AAAI writing scan.

The maintained teaching selection spans contrasting contribution styles and research areas. It is neither random nor statistically representative. Selected-study completion is separate from full-inventory coverage, and its measurements cannot establish conference-wide writing frequencies or a standard figure quota.

Main manuscripts use official OJS proceedings PDFs. Separately linked posters, code, extended versions and supplements are not counted as inspected appendices unless a study explicitly records a second inspected source. Formatting lessons use the 2026 call and author kit, not a later edition.

Inspect the complete inventory and source status · Official oral program

Caption labels, not panels. A six-panel figure counts once. Observed figures, tables and section spans describe the inspected source versions; they are not quotas or acceptance thresholds. For visual purpose, use the examples and type guides.

16 / 1058inspected PDFs / oral inventory records

144physical pages in those PDFs

7editorial contribution types

1separate files, excluded from totals

What the measured manuscripts look like

Observed extraction results, not a target length or figure quota. The middle 50% is the interval from the 25th to the 75th percentile. Missing boundaries are excluded from that span’s denominator, not replaced by zero. Caption counts are extraction-assisted and can miss nonstandard or image-only labels; inspect each paper’s notes.

Measure	Median	Middle 50%	Full range	Denominator
Inspected PDF lengthphysical pages	9	9–9	9–9	16 papers
Before references / appendixestimated page-equivalents	7.08	7.06–7.12	7.04–7.22	16 papers
Attached appendix spanestimated page-equivalents	Not resolved	Not resolved–Not resolved	Not resolved–Not resolved	0 papers
Main figurescaption labels	5	2.75–6	1–7	16 papers
Main tablescaption labels	3	1.75–4.25	0–8	16 papers
Appendix figurescaption labels	0	0–0	0–0	16 papers
Appendix tablescaption labels	0	0–0	0–0	16 papers
Separate source-file lengthwhole-file physical pages, possibly including repeated main text; not appendix-only length	13	13–13	13–13	1 papers
Separate source-file figureswhole-file caption labels; may include overlapping main-text displays	8	8–8	8–8	1 papers
Separate source-file tableswhole-file caption labels; may include overlapping main-text displays	10	10–10	10–10	1 papers

Compare like contributions

Each measured paper has one primary editorial type so the totals sum to 16. Secondary contributions were not separately coded; empty secondary lists do not imply the absence of mixed contributions. These are classifications of the annotated argument, not official AAAI tracks. The table uses the same extraction method and shows medians; the category guides below explain what these artifacts should accomplish.

Primary contribution	Papers	Main figures / tables	Body span ≈ pp.	Appendix span ≈ pp.
Architecture & representation	3	6 / 6	7.07	Not resolved0/3 attached boundaries; separate PDF median Not resolved pp. (0 papers)
Training, objectives & optimization	4	4 / 3	7.12	Not resolved0/4 attached boundaries; separate PDF median Not resolved pp. (0 papers)
Theory & formal analysis	2	3 / 1	7.06	Not resolved0/2 attached boundaries; separate PDF median Not resolved pp. (0 papers)
Empirical analysis, diagnosis & auditing	1	5 / 8	7.13	Not resolved0/1 attached boundaries; separate PDF median 13 pp. (1 papers)
Datasets, benchmarks & measurement	1	7 / 1	7.09	Not resolved0/1 attached boundaries; separate PDF median Not resolved pp. (0 papers)
Efficiency, inference & systems	2	4 / 2.5	7.14	Not resolved0/2 attached boundaries; separate PDF median Not resolved pp. (0 papers)
Domain-grounded methods & physical systems	3	4 / 3	7.06	Not resolved0/3 attached boundaries; separate PDF median Not resolved pp. (0 papers)

Search all 16 measurement records

Open a paper’s ledger for every detected top-level section, its estimated length and share of the PDF, and figure/table locations. Its existing study explains what those visuals and sections do.

Showing 16 of 16 records.

Paper / category	PDF pages	Main F / T	Appendix F / T	Body / appendix span ≈ pp.	Notes
CATP: Contextually Adaptive Token Pruning for Efficient and Enhanced Multimodal In-Context LearningEfficiency, inference & systems	9	6 / 3	0 / 0	7.07 / Unresolved	No attached appendix boundary detectedSource correction retained
ProGMLP: A Progressive Framework for GNN-to-MLP Knowledge Distillation with Efficient Trade-offsTraining, objectives & optimization	9	6 / 4	0 / 0	7.14 / Unresolved	No attached appendix boundary detectedSource correction retained
DeepProofLog: Efficient Proving in Deep Stochastic Logic ProgramsEfficiency, inference & systems	9	2 / 2	0 / 0	7.22 / Unresolved	No attached appendix boundary detectedSource correction retained
VCapsBench: A Large-scale Fine-grained Benchmark for Video Caption Quality EvaluationDatasets, benchmarks & measurement	9	7 / 1	0 / 0	7.09 / Unresolved	No attached appendix boundary detectedSource correction retained
FoundationSLAM: Unleashing the Power of Depth Foundation Models for End-to-End Dense Visual SLAMArchitecture & representation	9	6 / 6	0 / 0	7.07 / Unresolved	No attached appendix boundary detectedSource correction retained
Modality-Balanced Collaborative Distillation for Multi-Modal Domain GeneralizationTraining, objectives & optimization	9	6 / 5	0 / 0	7.05 / Unresolved	No attached appendix boundary detectedSource correction retained
Revisiting (Un)Fairness in Recourse by Minimizing Worst-Case Social BurdenTraining, objectives & optimization	9	2 / 2	0 / 0	7.22 / Unresolved	No attached appendix boundary detectedSource correction retained
Deep (Predictive) Discounted Counterfactual Regret MinimizationTraining, objectives & optimization	9	2 / 0	0 / 0	7.09 / Unresolved	No attached appendix boundary detectedSource correction retained
FoAM: Foresight-Augmented Multi-Task Imitation Policy for Robotic ManipulationDomain-grounded methods & physical systems	9	4 / 3	0 / 0	7.04 / Unresolved	No attached appendix boundary detectedSource correction retained
URaG: Unified Retrieval and Generation in Multimodal LLMs for Efficient Long Document UnderstandingArchitecture & representation	9	3 / 7	0 / 0	7.08 / Unresolved	No attached appendix boundary detectedSource correction retained
VAEVQ: Enhancing Discrete Visual Tokenization Through Variational ModelingArchitecture & representation	9	7 / 3	0 / 0	7.04 / Unresolved	No attached appendix boundary detectedSource correction retained
Why Do Open-Source LLMs Struggle with Data Analysis? A Systematic Empirical StudyEmpirical analysis, diagnosis & auditing	9	5 / 8	0 / 0	7.13 / Unresolved	No attached appendix boundary detectedSource correction retainedSeparate files: 13 pp.; 8 figures / 10 tables; excluded from main totals
Graph-Semantic Guided Learning for Virtual Immunohistochemistry Staining on Consecutive Histology SectionsDomain-grounded methods & physical systems	9	4 / 4	0 / 0	7.06 / Unresolved	No attached appendix boundary detectedSource correction retained
MSAnchor: De Novo Molecular Generation from Mass Spectrometry Data with Anchor-Extended Molecular ScaffoldsDomain-grounded methods & physical systems	9	6 / 2	0 / 0	7.11 / Unresolved	No attached appendix boundary detectedSource correction retained
On the Learning Dynamics of Two-layer Linear Networks with Label Noise SGDTheory & formal analysis	9	5 / 1	0 / 0	7.04 / Unresolved	No attached appendix boundary detectedSource correction retained
Generalization Bounds for Semi-supervised Matrix Completion with Distributional Side InformationTheory & formal analysis	9	1 / 1	0 / 0	7.09 / Unresolved	No attached appendix boundary detectedSource correction retained

Measurement method

The atlas measures only the source-inspected teaching collection, not all 1,058 oral papers. Uninspected records have no fabricated zero counts, section estimates or contribution classifications.

AAAI commonly uses unnumbered top-level headings. Main-section locators are source-bound corrections checked against the manuscript, with column-aware physical-page coordinates; lower-level bold lead-ins are not automatically counted as sections.

Figure and table counts refer to distinct caption labels, not subpanels or plot types. Examples below are selected editorial teaching material; observed counts are not a required display package.

Section spans run from one detected top-level heading to the next in physical-page coordinates. They include whitespace and floats, not exact text lengths, occupied-ink area or official page-limit accounting. Their percentages use the entire inspected PDF; the title, abstract and teaser can precede the first heading.

Missing boundaries remain unresolved rather than zero. Separately inspected files have separate denominators. Caption counts are extraction-assisted; contents entries, repeated labels, references and image-only captions need source checks. Corrections remain associated with the exact source URL.

Contribution labels are editorial, not official tracks. Selected examples do not measure visual-style prevalence, and accepted-paper observations cannot show which writing choices caused acceptance.

Download structural metadata · Inspect individual sources and studies