MMLoSo 2025
Multimodal Models for Low-Resource Contexts and Social Impact
📍 Co-located with IJCNLP-AACL 2025
📅 23rd December 2025, Mumbai, India
📖 About
This workshop brings together researchers at the intersection of multimodal learning, NLP, and AI for social good, with a focus on low-resource and underserved settings… We invite papers on developing robust and inclusive multimodal systems that can operate effectively under data constraints. Topics include learning with multiple modalities, cross-lingual, cross-modal adaptation, and interpretable models for domains like healthcare, ecological monitoring, education, and cultural heritage preservation. Alongside papers and keynotes, a community-driven shared task will evaluate multimodal robustness and generalization in low-resource contexts.
📢 Call for Papers
We focus on bridging the gap between the growing capabilities of multimodal machine learning models and the urgent needs of real-world applications in under-resourced, marginalized, or data-constrained settings. This includes scenarios where data is scarce, modalities are incomplete or imbalanced, and computational or human infrastructure may be limited.
The topics of interest for the workshop include, but are not limited to:
-
Learning with Missing or Incomplete Modalities
Techniques for modality dropout, hallucination, and imputation when input signals are sparse or missing at training or inference time. -
Few-Shot, Zero-Shot, and Transfer Learning in Multimodal Contexts
Approaches that allow models pre-trained on high-resource datasets to adapt effectively to novel, low-resource domains and languages. -
Multilingual and Multimodal Representation Learning
Unifying language, vision, audio, and other modalities across multiple languages, especially those underrepresented in current benchmarks. -
Ethical, Interpretable, and Responsible AI for Multimodal Systems
Auditing and mitigating bias in multimodal systems; developing transparent models that explain decisions across modalities in high-stakes domains. -
Benchmarking and Evaluation for Real-World Robustness
Proposing new datasets, metrics, and evaluations that reflect deployment challenges in regions with limited resources or infrastructure. -
Applications in Social Good, including:
- Ecological and biodiversity monitoring (e.g., combining satellite, image, and audio data)
- Public health and epidemiology in underserved regions
- Language documentation and cultural preservation
- Crisis response, misinformation detection, and social justice
📝 Submission Guidelines
The workshop invites submissions in two formats. All page limits exclude unlimited pages for references.
- Long papers (up to 8 pages): This format is for presenting a substantial and complete version of research.
- Short papers (up to 4 pages): This format is ideal for presenting more focused contributions, including preliminary work, work-in-progress, position papers, or descriptions of systems.
📅 Submission link: OpenReview Submission Portal
📑 Review process: Review Process: Each paper will be reviewed by at least three program committee members. The submissions will be evaluated based on their technical soundness, relevance, significance, originality, and clarity.
🏆 Shared Task & Competition
Full details are on the dedicated page.
🎤 Keynote Speakers
Details coming soon..
📅 Important Dates
(Subject to change based on IJCNLP-AACL 2025)
- 📝 Submission Deadline: September 29, 2025
- 📢 Notification: November 3, 2025
- 🖋 Camera-ready Deadline: November 11, 2025
- 📍 Workshop: December 23, 2025
👥 Organizers
Ankita Shukla
University of Nevada, Reno
Sandeep Kumar
IIT Delhi
Amrit Singh Bedi
University of Central Florida
Tanmoy Chakraborty
IIT Delhi
📝 Program Committee
We’re looking for dedicated reviewers to help maintain the high standard of our review process. Each reviewer is assigned no more than two submissions. Interested? Sign up here
🙋♀️ Student Volunteers
- Shashwat Bhardwaj
- Debarchan Basu
- Vaibhav Sharma
-
Pooja Singh
✉️ Contact
For any queries, email us at:
📧 ankitas@unr.edu