EffL Lab

Make AI accessible to everyone with Efficient ML!

Efficient Learning Lab, or simply EffL, is a research group led by Jaeho Lee.
We develop theories, algorithms, and systems to make ML more efficient.

To join us, apply us through POSTECH EE or Graduate School of AI.
For details, see here.

News #

Two papers will be presented at CVPR 2024 🎊.

  • An intriguing way to accelerate neural field training (Oral, led by Junwon and Sangyoon).

  • A way to characterize visual biases in the form of text keywords (Highlight, collaboration with UMich and KAIST).



A paper on a distillation-based method to recognized named entities in multi-modal setup will be presented at NAACL 2024! 🎊

This is a joint work with NAVER.


A paper on unlearning will be presented at IEEE S&P 2024.


One conference paper and two workshop papers will be presented.

  • Conference: Large-scale meta-learning for neural fields (with Google DeepMind).
  • Workshop: Model merging (led by Jiwoon) and Neural field training (led by Junwon and Sangyoon).


Jaeho and Seungwoo started working at Google.

They will work on LLM optimization as a visiting faculty and a student intern, respectively.


One conference paper and three workshop papers has been presented.

  • Conference: Data compression algorithm (with Google DeepMind).
  • Workshop: Two on spurious correlations, and one on large-batch training.


A paper on spurious correlation of conditional generative model has been published.

This is a joint work with Samsung Advanced Institute of Technology.