Opportunistic Expert Activation: Batch-Aware Expert Routing for Faster Decode Without Retraining

Documenti analoghi

• Seesaw: Accelerating Training by Balancing Learning Rate and Batch Size Scheduling
  di: Meterez, Alexandru, et al.
  Pubblicazione: (2025)
• Beat the long tail: Distribution-Aware Speculative Decoding for RL Training
  di: Shao, Zelei, et al.
  Pubblicazione: (2025)
• The Recurrent Transformer: Greater Effective Depth and Efficient Decoding
  di: Oncescu, Costin-Andrei, et al.
  Pubblicazione: (2026)
• Flash Inference: Near Linear Time Inference for Long Convolution Sequence Models and Beyond
  di: Oncescu, Costin-Andrei, et al.
  Pubblicazione: (2024)
• RouteHijack: Routing-Aware Attack on Mixture-of-Experts LLMs
  di: Xu, Zhiyuan, et al.
  Pubblicazione: (2026)