Closed-domain Question Answering on BoolQ
85.2EMVanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| VanillaBackbone=Mistral-7B-Instruct-v0.22026.05 | 85.2 | |
| LoRABackbone=Mistral-7B-Instruct-v0.22026.05 | 85 | |
| VanillaBackbone=Qwen2.5-7B-Instruct2026.05 | 84 | |
| LoRABackbone=Qwen2.5-7B-Instruct2026.05 | 84 | |
| SafeDecodingBackbone=Qwen2.5-7B-Instruct2026.05 | 83.8 | |
| EVA (text)Backbone=Qwen2.5-7B-Instruct2026.05 | 83.2 | |
| LEDBackbone=Qwen2.5-7B-Instruct2026.05 | 81.9 | |
| EVA (text)Backbone=Mistral-7B-Instruct-v0.22026.05 | 81.4 | |
| Circuit BreakersBackbone=Mistral-7B-Instruct-v0.22026.05 | 81.1 | |
| LEDBackbone=Mistral-7B-Instruct-v0.22026.05 | 80.3 | |
| Circuit BreakersBackbone=Qwen2.5-7B-Instruct2026.05 | 79.7 | |
| LoRABackbone=Llama3.1-8B-Instruct2026.05 | 78.2 | |
| VanillaBackbone=Vicuna-7B-v1.52026.05 | 77.7 | |
| VanillaBackbone=Llama3.1-8B-Instruct2026.05 | 77 | |
| LoRABackbone=Llama2-7B-chat2026.05 | 76.9 | |
| LEDBackbone=Llama3.1-8B-Instruct2026.05 | 76.8 | |
| EVA (text)Backbone=Vicuna-7B-v1.52026.05 | 76.2 | |
| LEDBackbone=Vicuna-7B-v1.52026.05 | 76 | |
| Circuit BreakersBackbone=Vicuna-7B-v1.52026.05 | 75.6 | |
| EVA (text)Backbone=Llama3.1-8B-Instruct2026.05 | 75.2 | |
| SafeDecodingBackbone=Llama3.1-8B-Instruct2026.05 | 74.6 | |
| LoRABackbone=Vicuna-7B-v1.52026.05 | 74.2 | |
| VanillaBackbone=Llama2-7B-chat2026.05 | 73.4 | |
| SafeDecodingBackbone=Mistral-7B-Instruct-v0.22026.05 | 73.2 | |
| EVA (text)Backbone=Llama2-7B-chat2026.05 | 71.8 | |
| Circuit BreakersBackbone=Llama2-7B-chat2026.05 | 71.2 | |
| LEDBackbone=Llama2-7B-chat2026.05 | 70.5 | |
| SafeDecodingBackbone=Llama2-7B-chat2026.05 | 68.8 | |
| SafeDecodingBackbone=Vicuna-7B-v1.52026.05 | 67.1 | |
| Circuit BreakersBackbone=Llama3.1-8B-Instruct2026.05 | 48.9 |