aiqbench analysis
SSA addresses the performance degradation and capability gaps inherent in sparse attention by aligning full and sparse attention outputs in feature space, effectively mitigating distribution mismatch and incomplete gradient flow. Theoretical analysis confirms that approximation error scales linearly with dropped attention mass, while experiments show the method achieves state-of-the-art results across varying sparsity budgets and long-context scenarios. This approach offers a robust framework for integrating efficient sparse mechanisms without sacrificing the performance of full-attention models.
SSA: Sparse Sparse Attention by Aligning Full and Sparse Attention Outputs in Feature Space
arXiv:2511.20102v3 Announce Type: replace Abstract: Sparse attention reduces the quadratic complexity of full self-attention but faces two challenges: (1) an attention gap, where applying sparse attention to full-attention-trained models causes performance degradation due to train-inference distribution mismatch, and (2) a capability gap, where models trained purely with sparse attention lack complete gradient flow, preventing them from matching full-attention performance. We propose SSA (Sparse Sparse Attention), a training framework that integrates both sparse and full attention with bidirectional attention-output alignment. We prove that the approximation error scales linearly with the attention mass dropped under sparse attention, and show that SSA's alignment objective substantially reduces this quantity compared to baselines. Experiments demonstrate that SSA achieves state-of-the-art performance under both inference modes, adapts smoothly to varying sparsity budgets, and demonstrates superior long-context capabilities.
Abstract:Sparse attention reduces the quadratic complexity of full self-attention but faces two challenges: (1) an attention gap, where applying sparse attention to full-attention-trained models causes performance degradation due to train-inference distribution mismatch, and (2) a capability gap, where models trained purely with sparse attention lack complete gradient flow, preventing them from matching full-attention performance. We propose SSA (Sparse Sparse Attention), a training framework that integrates both sparse and full attention with bidirectional attention-output alignment. We prove that the approximation error scales linearly with the attention mass dropped under sparse attention, and show that SSA's alignment objective substantially reduces this quantity compared to baselines. Experiments demonstrate that SSA achieves state-of-the-art performance under both inference modes, adapts smoothly to varying sparsity budgets, and demonstrates superior long-context capabilities.
Computation and Language (cs.CL)
arXiv:2511.20102 [cs.CL]
(or arXiv:2511.20102v3 [cs.CL] for this version)
Full text under the publisher's open licence · 137 words
Related coverage
SSA: Sparse Sparse Attention by Aligning Full and Sparse Attention Outputs in Feature Space
arXiv · Jun 5, 2026
Selective Coupling of Decoupled Informative Regions: Masked Attention Alignment for Data-Free Quantization of Vision Transformers
arXiv · Jun 4, 2026
Selective Coupling of Decoupled Informative Regions: Masked Attention Alignment for Data-Free Quantization of Vision Transformers
arXiv · Jun 8, 2026
Self-Explaining Hate Speech Detection with Moral Rationales
arXiv · Jul 21, 2026