Question Answering on PopQA (EM, F1, Retrieval Rate, Latency)
23EMMargin
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MarginBackbone=Qwen2.5-7B-Instruct, tau=0.352025.11 | 23 | 23.1 | 12.4 | 1.761 | |
| VarianceBackbone=Qwen2.5-7B-Instruct, tau=0.402025.11 | 22.8 | 22.9 | 18.2 | 1.847 | |
| EntropyBackbone=Qwen2.5-7B-Instruct, tau=0.802025.11 | 22.4 | 22.3 | 12.4 | 1.761 | |
| NeverBackbone=Qwen2.5-7B-Instruct2025.11 | 20 | 20.1 | 0 | 2.129 | |
| AlwaysBackbone=Qwen2.5-7B-Instruct2025.11 | 14.6 | 14.6 | 100 | 3.828 |