Question Answering on HotpotQA (test) (LLM-as-a-Judge)
0.77LLM Judge ScorePersonalAI
Evaluation Results
| Method | Links | |
|---|---|---|
| PersonalAILLM=GPT4o-mini, Retrieval algorithm=BeamSearch + WaterCircles, Graph restriction=no restrictions2025.06 | 0.77 | |
| PersonalAILLM=DeepSeek V3, Retrieval algorithm=BeamSearch + WaterCircles, Graph restriction=thesis vertices excluded2025.06 | 0.76 | |
| PersonalAILLM=Llama3.1 8B, Retrieval algorithm=BeamSearch, Graph restriction=object vertices excluded2025.06 | 0.47 | |
| PersonalAILLM=Qwen2.5 7B, Retrieval algorithm=BeamSearch, Graph restriction=object vertices excluded2025.06 | 0.24 | |
| PersonalAILLM=DeepSeek R1 7B, Retrieval algorithm=BeamSearch, Graph restriction=object vertices excluded2025.06 | 0.19 |