New frameworks tackle generalized category discovery under domain shifts

Researchers have developed three new frameworks to improve Generalized Category Discovery (GCD) when dealing with domain shifts in data. These methods adapt existing foundation models, including vision and vision-language models, to better categorize unlabeled instances from both known and unknown classes. The proposed techniques, HiLo, HLPrompt, and VLPrompt, utilize feature disentanglement, prompt tuning, and cross-modal consistency to handle semantic and domain shifts effectively. Experiments show significant improvements over existing approaches on various multi-domain shift datasets. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces novel techniques for improving model generalization and categorization under challenging domain shifts.

RANK_REASON This is a research paper detailing new methods for Generalized Category Discovery. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• HiLo
• HLPrompt
• VLPrompt
• Generalized Category Discovery

• paper
• other