Lizhong Zheng
1 indexed paper
Recent (6 mo)
1With code
0Influential cites
0Benchmarked
0Publications per year
126
Top categories
ML×1AI×1Stats Apps×1Stats ML×1
Frequent co-authors
Research Timeline
2026
When Softmax Fails at the Top: Extreme Value Corrections for InfoNCE
The paper proposes WEINCE, a modified InfoNCE objective that uses extreme value theory corrections to improve contrastive learning by more accurately modeling the selection of hard negative examples.
Highlighted terms show continued research focus across papers
Papers
cs.LGcs.AIstat.APRecentMay 29, 2026
When Softmax Fails at the Top: Extreme Value Corrections for InfoNCE
Melihcan Erol, Suat Evren, Oktay Ozel, Alexander Morgan +2 more
The paper proposes WEINCE, a modified InfoNCE objective that uses extreme value theory corrections to improve contrastive learning by more accurately modeling the selection of hard negative examples.
View →