From TLinFormer to TConstFormer: The Leap to Constant-Time Transformer Attention: Achieving O(1) Computation and O(1) KV Cache during Autoregressive Inference

Documenti analoghi

• Rethinking Transformer Connectivity: TLinFormer, A Path to Exact, Full Context-Aware Linear Attention
  di: Tang, Zhongpan
  Pubblicazione: (2025)
• CacheFormer: High Attention-Based Segment Caching
  di: Singh, Sushant, et al.
  Pubblicazione: (2025)
• TimeFormer: Transformer with Attention Modulation Empowered by Temporal Characteristics for Time Series Forecasting
  di: Liu, Zhipeng, et al.
  Pubblicazione: (2025)
• Compression is Routing: Reconstruction Error as an Intrinsic Signal for Modular Language Models
  di: Tang, Zhongpan
  Pubblicazione: (2025)
• SFi-Former: Sparse Flow Induced Attention for Graph Transformer
  di: Li, Zhonghao, et al.
  Pubblicazione: (2025)