CurvPrompt framework enhances dynamic graph prompting for low-label tasks

Researchers have developed CurvPrompt, a novel framework designed to improve dynamic graph prompting for tasks with limited labeled data. This method addresses the issue of "geometry under-adaptation" by dynamically routing nodes to a diverse set of Riemannian experts based on local topology and time. CurvPrompt utilizes soft routing during pre-training to establish a continuous mapping between topology and geometry, then transitions to hard Top-K routing for downstream adaptation, demonstrating significant advancements in few-shot link prediction and strong performance in node classification. AI

IMPACT This research could improve the efficiency and accuracy of machine learning models working with dynamic graph data, particularly in scenarios with limited training examples.

RANK_REASON The item is an academic paper detailing a new method for dynamic graph prompting. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• CORE Recommender
• CurvPrompt
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv Recommender
• Influence Flower
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• model release

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