Year: 2025
Type: conference-paper
Abstract: Image embedding models such as OpenAI's CLIP do not in- herently store any temporal information in the embeddings they generate. This paper argues that while such embeddings capture rich semantic relationships, they lack an innate ca- pacity to encode chronological details, requiring fine-tuning or additional layers to infer temporal attributes
Institution: University of Southern Maine
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4411549747