We propose a method for efficiently scaling up influence functions for LLMs of 100B parameters.
1 min read · October 05, 2025
2025