New framework enhances AI's long-video memory by tracking entity biographies

Researchers have developed a new framework called Grounded Entity Biographies (GEB) to improve long-video memory for AI systems. GEB addresses the challenge of resolving physical identity and connecting observations of the same object across extended periods. By grouping visually grounded instances into retrievable biographies, GEB allows models to track entities through events, leading to better performance on question-answering tasks. Evaluations on benchmarks, including day-long and week-long recordings, showed significant improvements, with GEB achieving 72.0% accuracy on the EgoLifeQA benchmark. AI

IMPACT Enhances AI's ability to understand and recall information from long videos, potentially improving applications in surveillance, content analysis, and personal assistants.

RANK_REASON The cluster contains a research paper detailing a new framework for AI memory. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• alphaXiv
• arXiv
• CatalyzeX
• CORE Recommender
• DagsHub
• EgoLifeQA
• Gotit.pub
• Grounded Entity Biographies
• Hugging Face
• Influence Flower
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →