Spring view of Zōjō-ji and Tokyo Tower with cherry blossoms
Conference on Parsimony and Learning (CPAL)
March 2027, Tokyo

Call for Papers — CPAL 2027

The Fourth Conference on Parsimony and Learning
March 23–26, 2027 · Hitotsubashi Hall, Tokyo, Japan

CPAL 2027 invites research that makes parsimony central to learning. Choose the Proceedings Track for original, unpublished archival papers or the Recent Spotlight Track for non-archival presentations of timely research. Below are the scientific scope, track requirements, deadlines, and review policies in one place.

Submission portal and templates Key Dates

On this page

About CPAL

Modern learning systems operate in enormous ambient spaces, yet their data, representations, dynamics, and solutions often admit much simpler descriptions. The Conference on Parsimony and Learning (CPAL) brings together researchers seeking to understand, discover, and exploit these structures across machine learning, signal processing, optimization, statistics, applied mathematics, neuroscience, scientific computing, and intelligent systems.

CPAL 2027 invites theoretical, methodological, empirical, and systems contributions in which parsimony is central. Parsimony may take the form of sparsity, low rank or intrinsic dimension, symmetry, modularity, compositionality, compressibility, simple mechanisms, or structured computation. Work on foundation, generative, multimodal, and agentic models is encouraged when the underlying parsimony principle is explicit. Authors should explain what the parsimonious structure is, whether it is assumed, learned, or emergent, and what it enables.

Subject Areas

CPAL 2027 welcomes contributions in which a clear parsimony principle is central. The following list is illustrative rather than exhaustive.

Theory and Foundations

  • Sparsity, structured sparsity, low rank, subspaces, manifolds, tensors, graphs, latent-variable models, and other low-dimensional structures.
  • Statistical and computational foundations of representation and feature learning.
  • Identifiability, generalization, robustness, implicit bias, and statistical–computational tradeoffs.
  • Symmetry, invariance, equivariance, causality, modularity, and compositional structure.
  • Information-theoretic, minimum-description-length, and compression-based views of learning.
  • Theoretical neuroscience, cognitive science, and biologically inspired mechanisms for parsimonious learning.

Methods and Models

  • Sparse coding, dictionary learning, matrix and tensor methods, inverse problems, and structured recovery.
  • Optimization and feature-learning dynamics that discover or induce parsimonious structure.
  • Interpretable and structured neural architectures, including algorithm unrolling.
  • Pruning, quantization, distillation, low-rank adaptation, modular networks, mixtures of experts, and conditional computation.
  • Data-, parameter-, memory-, energy-, and compute-efficient training and inference.
  • Parsimonious foundation, generative, multimodal, federated, continual, and agentic learning.

Systems, Data, and Applications

  • Hardware–algorithm and software–system co-design for structured or sparse computation.
  • Benchmarks, datasets, and metrics that reveal or evaluate parsimony and accuracy–efficiency tradeoffs.
  • Parsimonious learning for perception, action, reasoning, robotics, and control.
  • Scientific machine learning, signal and image processing, neuroscience, biology, medicine, engineering, and social science.
  • Resource-constrained, distributed, networked, and real-world intelligent systems.

Submissions proposing a new notion of parsimony are especially welcome when the connection is clearly motivated. Authors who are uncertain about fit may contact the Program Chairs at pcs@cpal.cc.

Submission Tracks

Feature Proceedings Track Recent Spotlight Track
Status Archival Non-archival presentation track
Review Double-blind Single-blind; do not anonymize
Format CPAL template; 9 pages of main text 250-word abstract plus supporting material
Prior/concurrent work No substantially similar prior or concurrent archival submission Concurrent, under-review, and recently published work permitted
Deadline See Proceedings dates See Recent Spotlight dates
Publication Proceedings of Machine Learning Research (PMLR) No proceedings or DOI
Presentation Poster; selected papers may receive talks Poster; selected contributions may receive short talks

Important Dates

Unless otherwise stated, submission deadlines are 23:59 Anywhere on Earth (AoE).

Proceedings dates

Event Date Countdown
Abstract registration November 23, 2026
Proceedings paper deadline December 5, 2026
Reviews released; author rebuttal begins January 11, 2027
Author rebuttal January 11–17, 2027
Author–reviewer discussion ends January 22, 2027
Proceedings notification February 1, 2027
Camera-ready deadline February 12, 2027

Author rebuttal runs January 11–17 (7 days), followed by author–reviewer discussion January 18–22 (5 days).

Recent Spotlight dates

Event Date Countdown
Recent Spotlight submission deadline January 18, 2027
Recent Spotlight notification February 1, 2027

See Key Dates for the conference, tutorial, and Rising Stars calendars.

Submission Requirements

Proceedings Track (archival)

The Proceedings Track is intended for original, unpublished research. The submission and review process is double-blind and will be hosted on OpenReview. Proceedings submissions must use the CPAL 2027 LaTeX style. The main text may contain up to nine pages, including figures and tables; references and appendices do not count toward the limit. The main text should be self-contained, and reviewers are not required to read appendices.

  • Submissions must be anonymized.
  • A public preprint does not violate the anonymity policy. Authors should refer to their own work in the third person where appropriate.
  • A submission must not be substantially similar to work already published, accepted, or simultaneously under review at another archival conference or journal.
  • Prior presentation at a non-archival workshop is permitted provided that the work did not appear in archival proceedings, a journal, or a book.

Recent Spotlight Track (non-archival)

The Recent Spotlight Track showcases timely research at different stages of development, from technically mature work in progress to recently accepted or published results. It is a presentation track and has no proceedings or DOI.

Each submission includes a 250-word abstract and one of the following:

  • a conference-style manuscript describing the work;
  • a poster PDF presenting work in progress; or
  • the camera-ready version of recently accepted or published work.

Authors may additionally provide appendices or public links that help explain the work. Concurrent or under-review submissions are permitted, as is work accepted or published at an archival venue within the 12 months preceding the Recent Spotlight deadline. Reviewing is single-blind, so authors should not anonymize their submissions.

Selection considers CPAL fit, clarity, technical credibility, timeliness, and discussion value, calibrated to the stated stage of the work. Authors remain responsible for complying with the policies of any other venue to which the same work is submitted.

Review Process

OpenReview hosts submissions, reviews, author responses, and discussion.

Review Model

  • Proceedings Track: double-blind review.
  • Recent Spotlight Track: single-blind review; authors do not anonymize their submissions.
  • Proceedings authors may respond to reviewers and update their manuscript during the published author–reviewer discussion period.
  • After author discussion, reviewers and Area Chairs discuss the submission and prepare recommendations.
  • Final decisions are made by the Program Chairs with Area-Chair recommendations and paper-level shepherding.

What CPAL Reviews For

Reviewers should assess whether a submission is:

  • technically sound;
  • clearly presented;
  • relevant to CPAL and explicit about its connection to parsimony;
  • appropriately supported by theory, experiments, or evidence for the claims being made; and
  • of genuine interest or discussion value to the CPAL research community.

CPAL does not require novelty for novelty’s sake or a new benchmark state of the art. A strong submission should make clear what new understanding, method, evidence, or scientific perspective the community gains from the work.

For Recent Spotlight submissions, judgments should be calibrated to the stated stage of the project. Work in progress is not expected to have the same degree of completion as an archival Proceedings submission, but it should still be technically credible, clear, timely, and valuable for discussion.

Reviewers should promptly report conflicts of interest or assignment problems through the OpenReview process. Questions about review policy may be directed to the Program Chairs at pcs@cpal.cc.

Presentation and Attendance

At least one author of every accepted Proceedings or Recent Spotlight contribution must register for CPAL 2027 and present the work in person. All accepted paper-track contributions are expected to be presented as posters; selected Proceedings papers and Recent Spotlight contributions may additionally be invited for oral presentation.

Use of AI-assisted tools

Authors may use AI-assisted tools in conducting research and preparing manuscripts, but remain fully responsible for the accuracy, originality, citations, ethics, and scientific integrity of all submitted content. Important, original, or non-standard methodological use of such tools should be disclosed sufficiently for readers to understand and reproduce the work. Routine spelling, grammar, formatting, or basic coding assistance need not be disclosed. AI systems should not be listed as authors.

Reviewers and committee members must preserve submission confidentiality and may not upload unpublished submission material to external AI systems.

Submission and Contact

Proceedings and Recent Spotlight submissions will be handled through OpenReview. The CPAL 2027 submission portal and LaTeX/Overleaf template links are collected on the CPAL OpenReview page as they become active.

For questions about scientific scope or submission policy, contact the Program Chairs at pcs@cpal.cc.

All participants are expected to follow the Code of Conduct.