Call for Papers

We invite submissions to the ACCV 2026 Workshop on Multimodal AI for Science and the Real World, to be held in person on December 15, 2026, at the Osaka International Convention Center (Grand Cube Osaka).

Scientific and real-world phenomena are rarely observed directly or completely. Instead, researchers must reason from partial, noisy, and heterogeneous measurements acquired through different instruments and modalities. This workshop seeks contributions that integrate such observations and connect them with scientific structure—including physics, chemistry, biology, geometry, and causal relationships—to recover latent mechanisms and support scientific discovery and real-world decision making.

We welcome contributions from computer vision, multimodal learning, scientific machine learning, sensing, foundation models, causal inference, world modeling, and relevant domain sciences.

Topic of Interest

We welcome submissions on (but not limited to):

• Multimodal learning from incomplete, noisy, and heterogeneous scientific observations
• Integration of vision with language, biological and omics signals, sensor readings, molecular data, time series, and simulation outputs
• Learning methods informed by physics, chemistry, biology, geometry, or causal structure
• Multimodal foundation models for scientific discovery and real-world decision making
• Vision-language, audio-language, sensor-language, and 3D-language models
• Multimodal reasoning over images, text, tables, equations, graphs, videos, and simulations
• Data-efficient and label-efficient learning for scientific domains with limited or noisy annotations
• Applications in medicine, biology, materials science, robotics, earth and environmental science, and other scientific or real-world domains

Accepted papers will be presented in person at the workshop. Presentation formats, including oral and poster presentations, will be determined by the organizers.

Important Dates

All deadlines are 23:59 Anywhere on Earth (AoE).

Submission Tracks

We offer two submission tracks.

Track 1: Archival Track

The archival track welcomes original, unpublished research. Submissions must not be under review at another archival venue during the review period.

Accepted papers will be included in the official ACCV 2026 workshop proceedings, subject to the ACCV publication, camera-ready, registration, and presentation requirements.

Submissions must follow the official ACCV 2026 paper format and will undergo double-blind peer review. Authors should carefully review the following instructions and resources before preparing their submissions:

• Author Guidelines
• ACCV 2026 Paper-format Example (PDF)
• Supplementary Material Requirements

The main paper must be self-contained. Supplementary material is optional and should provide supporting information rather than introduce essential content omitted from the main paper.

Track 2: Non-archival Poster Track

The non-archival track is intended for researchers who would like to present and discuss their work as a poster without archival publication.

This track will be administered separately through a Google Form rather than OpenReview. In accordance with guidance from the ACCV Workshop Chairs, OpenReview will be used for the archival track associated with the workshop proceedings. Since the non-archival track does not produce proceedings, it will use a streamlined screening process rather than a full archival peer-review workflow.

We welcome short papers and other concise contributions describing work in progress, preliminary results, ongoing projects, or recently completed research. Work that has already been accepted or published elsewhere is also welcome. Submissions that are under review or intended for submission elsewhere are eligible, provided that the authors comply with the policies of the other venue.

Formatting requirements depend on the status of the work at the time of submission:

• Work already accepted or published elsewhere: Authors may submit the accepted or published paper as-is in the original venue’s format; reformatting into the ACCV style is not required. This explicitly includes papers accepted at ACCV, CVPR, ICCV, ECCV, NeurIPS, ICLR, ICML, BMVC, and WACV. The four-page limit below does not apply to these submissions. The submission must identify the original venue and its acceptance or publication status, and authors must comply with that venue’s copyright and reuse policies.
• Work not yet accepted for publication: The submission, including work currently under review or intended for submission elsewhere, must follow the official ACCV 2026 paper format and contain no more than four pages, excluding references. Authors may also consult the paper-format example (PDF).

All submissions must:

• Be submitted as a PDF
• Provide sufficient information for the organizers to understand the motivation, approach, and main findings
• Include author names and affiliations; anonymization is not required

Authors are responsible for ensuring that they have the right to share the submitted PDF with the workshop organizers for screening and to present the work at the workshop.

Screening and Selection for the Non-archival Poster Track

• Papers already accepted at one of the major venues listed above will not be re-reviewed for technical quality. They will be checked only for eligibility, relevance to the workshop, compliance with the submission requirements, the authors’ right to share and present the work, and available poster capacity.
• Other submissions will receive a quick quality check based on relevance, clarity, technical soundness, and potential to stimulate discussion. This screening is not an archival peer-review process.
• If the number of eligible submissions exceeds the available poster capacity, the organizers may limit acceptances based on the criteria above and the available space.

Accepted submissions will:

• Be presented as posters at the workshop
• Not appear in the ACCV 2026 workshop proceedings
• Not be treated as archival publications
• Not be made publicly available by the workshop without the authors’ permission

Contact

Questions? Reach the organizers at mm4scireal@googlegroups.com.