Synthetic Data Generation for Multimodal Perception call for proposals - Fall 2026
Synthetic Data Generation for Multimodal Perception call for proposals - Fall 2026 supports researchers developing synthetic multimodal training data to help models understand everyday human activity.
Funder: Amazon Research Awards
Due Dates: October 1, 2026 (Submission opens) | November 4, 2026, 11:59 PM Pacific Time (Submission deadline)
Funding Amounts: Unrestricted funds averaging no more than $80,000 USD and AWS Promotional Credits averaging no more than $40,000 USD; one-year award period
Summary: Supports synthetic multimodal training data research for models that understand everyday human activity.
Key Information: PIs must be full-time faculty at accredited academic institutions or permanent researchers at legally recognized nonprofit NGOs.
Description
Amazon Research Awards, in connection with Amazon’s Devices and Services organization, supports research into synthetic training data for multimodal models that understand everyday human activity. Collecting large, diverse, densely annotated data from real environments is expensive, slow, and privacy-sensitive. This call seeks approaches that make broader activity coverage, causal consistency across modalities, longer coherent episodes, and scenario control scale with compute rather than human authoring.
The call welcomes theoretical advances, creative ideas, and practical applications across text-to-simulation motion and interaction synthesis, causally coherent audio-visual synthesis, long-horizon multimodal episodes, and action-conditioned world models. Proposals may address multiple topics. Priorities include physically plausible interactions, sounds grounded in visible events, persistent world state over hours-long episodes, and distinct plausible outcomes under different actions. Rigorous evaluation and interest in contributing open-source code, datasets, and benchmarks are central selection considerations.
