Research Paper
Original research papers including new techniques, position papers, literature surveys; maximum 9 pages of content plus unlimited references and appendix.
Computer Vision/CORE Unranked
MMRAgI 2026
The 2nd Workshop on Multi-Modal Reasoning for Agentic Intelligence
The 2nd Workshop on Multi-Modal Reasoning for Agentic Intelligence (MMRAgI) at CVPR 2026 focuses on advancing AI agents through Multimodal Foundation Models (MFMs) that integrate vision, language, and audio to enable richer perception and reasoning. The workshop aims to address challenges in cross-modal alignment, computational efficiency, causal reasoning, and real-world applications such as robotics, digital agents, and scientific AI, fostering inclusive dialogue among researchers from diverse backgrounds.
Key deadlines
Verified September 9, 2026Full paper
April 20, 2026
AoE
Workshop timeline
Submission and decisions
Full paperKey deadline
April 20, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Demo Paper
Technical reports for demo track; maximum 9 pages (same as research papers), with a link to a video, website, or code repository showcasing the demo.
Research areas in scope
01
Research Topics
Semantically consistent alignment across vision, language, and audio modalitiesNovel training paradigms to mitigate computational burden of multimodal systemsNovel model architectures for native multimodal reasoningQuantifying and enhancing causal reasoning capabilities of multimodal foundation modelsUnified metrics for evaluating cross-modal reasoning in open-ended scenariosIncentivizing reasoning native in multimodal representationsBalancing reliance on different multimodal signals for inferenceReducing computational demands from redundant modalities like images and videosInfluence of multimodal signals on LLM agent behaviorBenchmarks and datasets for evaluating interactivity, multimodal integration, and real-world performanceInteractive human-centric foundation models integrating human priors for perceptual, generative, and embodiment tasksInteractive simulation and adaptation systems driven by human-centric foundation modelsScalable multimodal integration in human-centric foundation models using vision, language, audio, and motionReal-world applications in social robotics, autonomous systems, and digital content creation
Policies worth checking twice
- Papers are limited to eight pages (including figures and tables; references excluded) in CVPR style.
- Submission undergoes double-blind review; authors must anonymize submissions and avoid identifying information.
- Workshop is non-archival; submissions will not result in proceedings and can be submitted elsewhere.
- Dual submissions are allowed: papers accepted at CVPR 2026 or under review at other venues (e.g., ICML 2026) are welcome.
- Reviewers must maintain confidentiality and report conflicts of interest immediately.
- Appendices are not required to be read by reviewers.
- Submissions must be anonymized and uploaded as a single PDF via OpenReview.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026