Search papers, labs, and topics across Lattice.
1
0
3
1
Forget re-sampling heuristics – HAMR uses meta-learning to dynamically re-weight training examples based on difficulty and class representation, leading to significant gains on imbalanced NLP tasks.
