GBLA extends kernelized linear attention with local causal mixing, key gating, and gated RMSNorm; a 1:2 hybrid with self-attention matches full bidirectional self-attention quality on Yandex Music data while delivering up to 8.2x speedup at length 32768.
Title resolution pending
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
representative citing papers
GenPAS unifies common data augmentation strategies for generative recommendation as special cases of a bias-controlled stochastic sampling process and demonstrates gains in accuracy, data efficiency, and parameter efficiency on benchmarks and industrial data.
citing papers explorer
-
Gated Bidirectional Linear Attention for Generative Retrieval
GBLA extends kernelized linear attention with local causal mixing, key gating, and gated RMSNorm; a 1:2 hybrid with self-attention matches full bidirectional self-attention quality on Yandex Music data while delivering up to 8.2x speedup at length 32768.
-
Sequential Data Augmentation for Generative Recommendation
GenPAS unifies common data augmentation strategies for generative recommendation as special cases of a bias-controlled stochastic sampling process and demonstrates gains in accuracy, data efficiency, and parameter efficiency on benchmarks and industrial data.
- SIDInspector: A Mapping-First Diagnostic Resource for Semantic-ID Tokenizers