PiKE: Adaptive Data Mixing for Large-Scale Multi-Task Learning Under Low Gradient Conflicts

NeurIPSSpotlight2025

Authors: Zeman Li, Yuan Deng, Peilin Zhong, Meisam Razaviyayn, Vahab Mirrokni
Venue: NeurIPS 2025
Track: Spotlight

TL;DR

Modern foundation models are trained on diverse datasets to enhance generalization across tasks and domains…

Opening excerpt from the authors’ abstract. source

Read the paper

OpenReview

Topics

generalization dataset

← All NeurIPS 2025 Spotlight papers · Browse the whole archive