List your company

Question Answering on NQ-Open (EM, F1, Retrieval, Latency)

57.6EMMargin
14.23225.49136.7548.009Mar 7, 2025Apr 17, 2025May 29, 2025Jul 10, 2025Aug 20, 2025Oct 1, 2025Nov 12, 2025

Evaluation results

MethodLinks
MarginBackbone=Llama-3.1-8B-Instruct, tau=0.502025.1157.654.70.80.012
VarianceBackbone=Llama-3.1-8B-Instruct, tau=0.602025.1156.853.72.60.059
EntropyBackbone=Llama-3.1-8B-Instruct, tau=0.952025.1155.453.113.20.175
NeverBackbone=Llama-3.1-8B-Instruct2025.1153.851.7010.299
AlwaysBackbone=Llama-3.1-8B-Instruct2025.1148.646.11001.248
EntropyBackbone=Qwen2.5-7B-Instruct, tau=0.852025.1139.639.14.60.964
MarginBackbone=Qwen2.5-7B-Instruct, tau=0.202025.1139.638.830.41.295
NeverBackbone=Qwen2.5-7B-Instruct2025.1138.837.703.293
VarianceBackbone=Qwen2.5-7B-Instruct, tau=0.602025.1138.6381.20.291
AlwaysBackbone=Qwen2.5-7B-Instruct2025.1137.436.71002.922
Expanded DropBackbone=DeepSeek-V2-Lite-Chat, Capacity Factor=1.02025.0326.8———
BaselineBackbone=DeepSeek-V2-Lite-Chat, Capacity Factor=1.02025.0326.5———
Token DropBackbone=DeepSeek-V2-Lite-Chat, Capacity Factor=1.02025.0325.2———
BaselineBackbone=Qwen1.5-MoE-Chat, Capacity Factor=1.02025.0322.6———
Expanded DropBackbone=Qwen1.5-MoE-Chat, Capacity Factor=1.02025.0322.4———
Token DropBackbone=Qwen1.5-MoE-Chat, Capacity Factor=1.02025.0321.4———
BaselineBackbone=OLMoE-Instruct, Capacity Factor=1.02025.0317.5———
Expanded DropBackbone=OLMoE-Instruct, Capacity Factor=1.02025.0317———
Token DropBackbone=OLMoE-Instruct, Capacity Factor=1.02025.0315.9———