Picture for Alvin Tsang

Alvin Tsang

Large-scale cloze evaluation reveals that token prediction tasks are neither lexically nor semantically aligned

Add code
Oct 15, 2024
Viaarxiv icon