FreqDiff framework enhances Temporal Knowledge Graph extrapolation

Researchers have introduced FreqDiff, a novel Frequency-aware Diffusion framework designed to improve Temporal Knowledge Graph (TKG) extrapolation. This method addresses limitations in existing diffusion-based approaches by better distinguishing query-specific evidence from non-salient historical facts. FreqDiff formulates future fact prediction as a query-slot denoising process, incorporating a dual-stream denoiser that combines temporal dependency modeling with context-aware spectral calibration. Experiments on public benchmarks show FreqDiff achieving state-of-the-art performance. AI

IMPACT This new framework could improve the accuracy of predicting future facts from historical data in knowledge graphs.

RANK_REASON The cluster contains a research paper detailing a new framework for Temporal Knowledge Graph extrapolation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Connected Papers
• CORE Recommender
• FreqDiff
• Hugging Face
• Influence Flower
• Litmaps
• scite Smart Citations
• Temporal Knowledge Graph

• paper
• other

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