ResearchBenchmarksBias Evaluation on WinoGenderFollow0.068EBSDefault0.031960.275230.51850.76177Oct 29, 2024Evaluation ResultsMethodMethodLinksEBSDefaultBackbone=GPT-2 XLBackbone=GPT-2 XL2024.100.068ATLASBackbone=GPT-2 XLBackbone=GPT-2 XL2024.100.153DefaultBackbone=LLAMA 3Backbone=LLAMA 32024.100.255DefaultBackbone=GPT-JBackbone=GPT-J2024.100.37ATLASBackbone=LLAMA 3Backbone=LLAMA 32024.100.409DefaultBackbone=LLAMA 2Backbone=LLAMA 22024.100.728ATLASBackbone=LLAMA 2Backbone=LLAMA 22024.100.815ATLASBackbone=GPT-JBackbone=GPT-J2024.100.969