Siyan Zhao, Daniel Israel, Guy Van den Broeck, and Aditya Grover identify that standard padding-based prefilling in large language models wastes significant computation when batches contain prompts of varying lengths.

Authors

• Person: Samuel Tesfazgi, Leonhard Sprandl, Sandra Hirche Organization: AISTATS
  Track: Poster Session 3 - aistats 2026

Sources

• Track: Poster Session 3 - aistats 2026 virtual.aistats.org Samuel Tesfazgi, Leonhard Sprandl, Sandra Hirche · AISTATS via serper

Referenced by nodes (1)

• Large Language Models concept