Long Form Question Answering on ELI5
31.15ROUGE-LInSemRAG
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| InSemRAGLLM=GPT2026.05 | 31.15 | 35.42 | 13.45 | — | — | |
| InSemRAGLLM=DeepSeek2026.05 | 30.85 | 34.85 | 13.15 | — | — | |
| SELF-RAGLLM=GPT2026.05 | 30.2 | 34.09 | 12.83 | — | — | |
| SELF-RAGLLM=DeepSeek2026.05 | 29.6 | 33.41 | 12.57 | — | — | |
| InSemRAGLLM=Qwen2026.05 | 28.95 | 32.85 | 12.45 | — | — | |
| SELF-RAGLLM=Qwen2026.05 | 27.78 | 31.36 | 11.8 | — | — | |
| FiDLLM=GPT2026.05 | 27.37 | 34.72 | 10.52 | — | — | |
| IRCoTLLM=GPT2026.05 | 27.34 | 31.28 | 9.63 | — | — | |
| Wikipedia GraphLLM=GPT2026.05 | 26.91 | 32.01 | 9.15 | — | — | |
| FiDLLM=DeepSeek2026.05 | 26.82 | 34.03 | 10.31 | — | — | |
| IRCoTLLM=DeepSeek2026.05 | 26.79 | 30.65 | 9.44 | — | — | |
| Wikipedia GraphLLM=DeepSeek2026.05 | 26.37 | 31.37 | 8.97 | — | — | |
| SearChainInformation Retrieval=true2023.04 | 25.57 | — | — | — | — | |
| CoT-on-ConcatLLM=GPT2026.05 | 25.31 | 31.2 | 8.39 | — | — | |
| FiDLLM=Qwen2026.05 | 25.18 | 31.94 | 9.68 | — | — | |
| IRCoTLLM=Qwen2026.05 | 25.15 | 28.78 | 8.86 | — | — | |
| SearChain w/o CompletionInformation Retrieval=true, Completion=false2023.04 | 25.02 | — | — | — | — | |
| ReActLLM=GPT2026.05 | 24.81 | 30.21 | 7.26 | — | — | |
| CoT-on-ConcatLLM=DeepSeek2026.05 | 24.8 | 30.58 | 8.22 | — | — | |
| Wikipedia GraphLLM=Qwen2026.05 | 24.76 | 29.45 | 8.42 | — | — | |
| Plan-and-Solve w/ IRInformation Retrieval=true2023.04 | 24.56 | — | — | — | — | |
| Naïve RAGLLM=GPT2026.05 | 24.32 | 30.21 | 6.77 | — | — | |
| ReActLLM=DeepSeek2026.05 | 24.31 | 29.61 | 7.11 | — | — | |
| Tree-of-Thought w/ IRInformation Retrieval=true2023.04 | 24.2 | — | — | — | — | |
| React -> CoT-SCInformation Retrieval=true2023.04 | 24.05 | — | — | — | — | |
| Naïve RAGLLM=DeepSeek2026.05 | 23.83 | 29.61 | 6.63 | — | — | |
| Verify-and-EditInformation Retrieval=true2023.04 | 23.8 | — | — | — | — | |
| Multi-Hop RAGLLM=GPT2026.05 | 23.61 | 27.86 | 6.32 | — | — | |
| DSPInformation Retrieval=true2023.04 | 23.46 | — | — | — | — | |
| Direct RetrievalInformation Retrieval=true2023.04 | 23.4 | — | — | — | — | |
| CoT-on-ConcatLLM=Qwen2026.05 | 23.29 | 28.7 | 7.72 | — | — | |
| Self-AskInformation Retrieval=true2023.04 | 23.25 | — | — | — | — | |
| Multi-Hop RAGLLM=DeepSeek2026.05 | 23.14 | 27.3 | 6.19 | — | — | |
| ToolFormerInformation Retrieval=true2023.04 | 23.05 | — | — | — | — | |
| SearChain w/o VerificationInformation Retrieval=true, Verification=false2023.04 | 22.98 | — | — | — | — | |
| ReActLLM=Qwen2026.05 | 22.83 | 27.79 | 6.68 | — | — | |
| SearChain w/o IRInformation Retrieval=false2023.04 | 22.54 | — | — | — | — | |
| Naïve RAGLLM=Qwen2026.05 | 22.37 | 27.79 | 6.23 | — | — | |
| Plan-and-SolveInformation Retrieval=false2023.04 | 22.23 | — | — | — | — | |
| Recite-and-answerInformation Retrieval=false2023.04 | 22.1 | — | — | — | — | |
| COT-SCInformation Retrieval=false2023.04 | 22.05 | — | — | — | — | |
| Least-to-MostInformation Retrieval=false2023.04 | 21.95 | — | — | — | — | |
| Direct PromptingInformation Retrieval=false2023.04 | 21.9 | — | — | — | — | |
| CoTInformation Retrieval=false2023.04 | 21.79 | — | — | — | — | |
| Self-Ask w/o IRInformation Retrieval=false2023.04 | 21.73 | — | — | — | — | |
| Multi-Hop RAGLLM=Qwen2026.05 | 21.72 | 25.63 | 5.81 | — | — | |
| Auto-CoTInformation Retrieval=false2023.04 | 21.55 | — | — | — | — | |
| w/o RAGLLM=GPT2026.05 | 20.51 | 23.62 | 4.25 | — | — | |
| w/o RAGLLM=DeepSeek2026.05 | 20.1 | 23.15 | 4.16 | — | — | |
| w/o RAGLLM=Qwen2026.05 | 18.87 | 21.73 | 3.91 | — | — | |
| InterleaveMultitask=true2019.11 | 14.63 | 23.32 | 4.79 | 30.8 | — | |
| SequentialMultitask=true2019.11 | 14.52 | 23.02 | 4.8 | 30.63 | — | |
| InterleaveMultitask=false2019.11 | 14.41 | 23.13 | 4.62 | 36.1 | — | |
| Multitask2019.11 | 14.35 | 22.48 | 4.63 | 31.66 | — | |
| Concatenate2019.11 | 14.33 | 22.86 | 4.54 | 36.37 | — | |
| Alternate2019.11 | 14.24 | 22.79 | 4.54 | 36.95 | — | |
| SequentialMultitask=false2019.11 | 13.83 | 22.14 | 4.24 | 50.23 | — | |
| DA-RAGBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 66.2 | |
| DA-RAGBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 69 | |
| DA-RAGBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 66.1 | |
| DRAGINBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 70.1 | |
| DRAGINBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 73.1 | |
| DRAGINBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 70.2 | |
| FLAREBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 69.8 | |
| FLAREBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 72.7 | |
| FLAREBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 69.8 | |
| GPT-4o2026.04 | — | — | — | — | 76.9 | |
| GuarantRAGBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 74.6 | |
| GuarantRAGBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 78.3 | |
| GuarantRAGBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 74.8 | |
| o12026.04 | — | — | — | — | 78.4 | |
| P-RAGBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 72.1 | |
| P-RAGBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 75.8 | |
| P-RAGBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 72.5 | |
| Qwen3-14BBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 59.2 | |
| Qwen3-14B-TBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 57.4 | |
| Qwen3-8BBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 57.1 | |
| RQ-RAGBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 68.7 | |
| RQ-RAGBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 71.8 | |
| RQ-RAGBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 68.9 | |
| SelfRAGBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 67.4 | |
| SelfRAGBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 70.1 | |
| SelfRAGBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 67.2 | |
| SOLARBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 70.2 | |
| SOLARBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 73.2 | |
| SOLARBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 70.6 | |
| Standard RAGBase Model=Qwen3-8B, Thinking Mode=false2026.04 | — | — | — | — | 61.5 | |
| Standard RAGBase Model=Qwen3-14B, Thinking Mode=false2026.04 | — | — | — | — | 64.3 | |
| Standard RAGBase Model=Qwen3-14B, Thinking Mode=true2026.04 | — | — | — | — | 61.8 |