Although foundation models (FMs) claim to be powerful, their generalization ability significantly decreases when faced with distribution shifts, weak supervision, or malicious attacks in the open world. On the other hand, most domain generalization or adversarial fine-tuning methods are task-related or model-specific, ignoring the universality in practical applications and the transferability between FMs. This paper delves into the problem of generalizing FMs to the out-of-domain data. We propose a novel framework, the Object-Concept-Relation Triad (OCRT), that enables FMs to extract sparse, high-level concepts and intricate relational structures from raw visual inputs. The key idea is to bind objects in visual scenes and a set of object-centric representations through unsupervised decoupling and iterative refinement. To be specific, we project the object-centric representations onto a semantic concept space that the model can readily interpret and estimate their importance to filter out irrelevant elements. Then, a concept-based graph, which has a flexible degree, is constructed to incorporate the set of concepts and their corresponding importance, enabling the extraction of high-order factors from informative concepts and facilitating relational reasoning among these concepts. Extensive experiments demonstrate that OCRT can substantially boost the generalizability and robustness of SAM and CLIP across multiple downstream tasks.

Biomedical engineers at Duke University have devised a method for increasing the resolution of optical coherence tomography (OCT) down to a single micrometer in all directions, even in a living patient. The new technique, called optical coherence refraction tomography (OCRT), could improve medical images obtained in the multibillion-dollar OCT industry for medical fields ranging from cardiology to oncology. The results appear in a paper published online on August 19 in the journal Nature Photonics. "An historic issue with OCT is that the depth resolution is typically several times better than the lateral resolution," said Joseph Izatt, the Michael J. Fitzpatrick Professor of Engineering at Duke. "If the layers of imaged tissues happen to be horizontal, then they're well defined in the scan. But to extend the full power of OCT for live imaging of tissues throughout the body, a method for overcoming the tradeoff between lateral resolution and depth of imaging was needed."