Question Answering on HalluQA
43.69AccuracyHaluSearch (MCTS)
Evaluation Results
| Method | Links | |
|---|---|---|
| HaluSearch (MCTS)Backbone=Qwen2-7B-Instruct2025.01 | 43.69 | |
| BoNBackbone=Qwen2-7B-Instruct2025.01 | 35.44 | |
| CoTBackbone=Qwen2-7B-Instruct2025.01 | 33.5 | |
| SCBackbone=Qwen2-7B-Instruct2025.01 | 32.52 | |
| Direct GenerationBackbone=Qwen2-7B-Instruct2025.01 | 32.04 | |
| Self-RefineBackbone=Qwen2-7B-Instruct2025.01 | 26.21 | |
| HaluSearch (MCTS)Backbone=Llama3.1-8B-Instruct2025.01 | 16.5 | |
| BoNBackbone=Llama3.1-8B-Instruct2025.01 | 12.62 | |
| Direct GenerationBackbone=Llama3.1-8B-Instruct2025.01 | 8.25 | |
| SCBackbone=Llama3.1-8B-Instruct2025.01 | 8.25 | |
| Self-RefineBackbone=Llama3.1-8B-Instruct2025.01 | 8.25 | |
| CoTBackbone=Llama3.1-8B-Instruct2025.01 | 6.8 |