List your company

Question Answering on ChineseFactEval

70.4AccuracyHaluSearch (MCTS)
27.96838.9845061.016Jan 2, 2025

Evaluation results

MethodLinks
HaluSearch (MCTS)Backbone=Qwen2-7B-Instruct2025.0170.4
BoNBackbone=Qwen2-7B-Instruct2025.0164
SCBackbone=Qwen2-7B-Instruct2025.0159.2
Direct GenerationBackbone=Qwen2-7B-Instruct2025.0156
CoTBackbone=Qwen2-7B-Instruct2025.0154.4
Self-RefineBackbone=Qwen2-7B-Instruct2025.0148.8
HaluSearch (MCTS)Backbone=Llama3.1-8B-Instruct2025.0140.8
SCBackbone=Llama3.1-8B-Instruct2025.0136
BoNBackbone=Llama3.1-8B-Instruct2025.0135.6
Self-RefineBackbone=Llama3.1-8B-Instruct2025.0133.6
CoTBackbone=Llama3.1-8B-Instruct2025.0132
Direct GenerationBackbone=Llama3.1-8B-Instruct2025.0129.6