glossary termAngestromTrust 60Published 1mo agoLive · 3mo ago
Attention
A mechanism that lets a model weigh which parts of the input matter most for each output.
A mechanism that lets a model weigh which parts of the input matter most for each output. A mechanism that lets a model weigh which parts of the input matter most for each output.
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- LinkedLinked via unknownlucidrains/poly-attention →
- LinkedLinked via unknownDnA: Denoising Attention for Visual Tasks →
- LinkedLinked via unknownattention-zoo →
- LinkedLinked via unknownBreakthrough in long-context efficiency announced →
- LinkedLinked via unknownSparse attention at million-token context →
- LinkedLinked via unknownPosition Bias Correction is Insufficient for One-Pass Attention Sorting →
- LinkedLinked via unknownMorphing into Hybrid Attention Models →
- LinkedLinked via unknownLogit-Contribution Scoring Identifies Non-Literal Retrieval Heads →
Related to (incoming)
repolucidrains/poly-attentionpaperDnA: Denoising Attention for Visual Tasksrepoattention-zoopaperSparse attention at million-token contextpaperPosition Bias Correction is Insufficient for One-Pass Attention SortingpaperMorphing into Hybrid Attention ModelspaperLogit-Contribution Scoring Identifies Non-Literal Retrieval HeadspaperInhibited Self-Attention: Sharpening Focus in Vision Transformers
Covers (incoming)
Related across the graph
paperPosition Bias Correction is Insufficient for One-Pass Attention SortingpaperInhibited Self-Attention: Sharpening Focus in Vision TransformerspaperLogit-Contribution Scoring Identifies Non-Literal Retrieval HeadspaperMorphing into Hybrid Attention ModelspaperSparse attention at million-token contextrepolucidrains/poly-attentionnewsBreakthrough in long-context efficiency announcednewsIntroducing DWARF-55M-BasepaperDnA: Denoising Attention for Visual Tasksrepoattention-zoo