Labelbox Labs introduces "Label Blocks," a new feature for multimodal data labeling that allows users to capture human feedback at higher abstraction levels, aligning with how humans make decisions in real-world contexts. As foundation models increasingly embrace multimodal capabilities, this feature facilitates the labeling of every data modality within a task at a granular level while enabling users to make comprehensive judgments about the task. This approach is particularly beneficial for professionals like medical practitioners who need to simultaneously consider various data types, such as patient histories and medical images, to make informed decisions. Labelbox Labs encourages users to explore a demo of Label Blocks and invites feedback from those interested in applying this innovation to their multimodal use cases, with an option to sign up for preview access.