Question Answering on TriviaQA (Context Relevance, Faithfulness, LLM-as-a-Judge)
80LLM-as-a-Judge ScorePAI-2
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PAI-2LLM=Qwen2.5 7B, Retrieval Mode=Combined Retrieval (best), Plan Enhancement=true, Algorithm Config=BS + NR / all2026.05 | 80 | 89 | 83 | |
| HippoRAG 2LLM=Qwen2.5 7B2026.05 | 77 | — | — | |
| PAI-2LLM=Qwen2.5 7B, Retrieval Mode=only Naive Retriever, Plan Enhancement=true2026.05 | 77 | 89 | 88 | |
| PAI-2LLM=Qwen2.5 7B, Retrieval Mode=only Traversal Algorithms, Plan Enhancement=true2026.05 | 77 | 92 | 91 | |
| RAPTORLLM=Qwen2.5 7B2026.05 | 73 | — | — | |
| PAI-1LLM=Qwen2.5 7B, Retrieval Mode=Combined Retrieval (best), Algorithm Config=BS+NR / all2026.05 | 73 | 66 | 82 | |
| PAI-2LLM=Qwen2.5 7B, Retrieval Mode=only Traversal Algorithms, Plan Enhancement=false2026.05 | 71 | 74 | 88 | |
| PAI-1LLM=Qwen2.5 7B, Retrieval Mode=only Naive Retriever2026.05 | 70 | 68 | 74 | |
| PAI-1LLM=Qwen2.5 7B, Retrieval Mode=only Traversal Algorithms2026.05 | 67 | — | — | |
| LightRAGLLM=Qwen2.5 7B2026.05 | 37 | — | — |