Open-Domain Question Answering on TriviaQA (SubEM, F1)
74.01SubEMStable-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Stable-RAGBackbone=LLaMA3-8B-Instruct, Retriever=DPR, exhaustive decoding=true2026.01 | 74.01 | 74.12 | |
| Stable-RAGBackbone=LLaMA3-8B-Instruct, Retriever=DPR, exhaustive decoding=false2026.01 | 73.43 | 73.76 | |
| Stable-RAGBackbone=LLaMA3-8B-Instruct, Retriever=Contriever, exhaustive decoding=true2026.01 | 72.13 | 71.89 | |
| Stable-RAGBackbone=LLaMA3-8B-Instruct, Retriever=Contriever, exhaustive decoding=false2026.01 | 72.05 | 71.56 | |
| Stable-RAGBackbone=Qwen3-8B, Retriever=DPR, exhaustive decoding=true2026.01 | 71.74 | 73.4 | |
| Stable-RAGBackbone=Qwen3-8B, Retriever=DPR, exhaustive decoding=false2026.01 | 71.32 | 72.89 | |
| ATMBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 70.12 | 70.35 | |
| ATMBackbone=Qwen3-8B, Retriever=DPR2026.01 | 70.06 | 71.67 | |
| Pos2DistillBackbone=Qwen3-8B, Retriever=DPR2026.01 | 69.87 | 71.35 | |
| Vanilla RAGBackbone=Qwen3-8B, Retriever=DPR2026.01 | 69.62 | 71.03 | |
| RetRobustBackbone=Qwen3-8B, Retriever=DPR2026.01 | 69.12 | 70.33 | |
| Ms-PoEBackbone=Qwen3-8B, Retriever=DPR2026.01 | 69.03 | 70.84 | |
| RetRobustBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 68.67 | 70.42 | |
| RAATBackbone=Qwen3-8B, Retriever=DPR2026.01 | 68.54 | 69.88 | |
| RAATBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 68.03 | 69.12 | |
| Vanilla RAGBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 67.12 | 68.61 | |
| Stable-RAGBackbone=Qwen3-8B, Retriever=Contriever, exhaustive decoding=true2026.01 | 67.11 | 68.79 | |
| Stable-RAGBackbone=Qwen3-8B, Retriever=Contriever, exhaustive decoding=false2026.01 | 66.58 | 68.13 | |
| Pos2DistillBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 66.57 | 68.12 | |
| Ms-PoEBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 66.48 | 67.73 | |
| ATMBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 66.37 | 67.12 | |
| RAATBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 65.58 | 66.94 | |
| Pos2DistillBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 64.95 | 66.81 | |
| Ms-PoEBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 64.88 | 66.72 | |
| RetRobustBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 64.85 | 66.72 | |
| ATMBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 64.78 | 66.57 | |
| Vanilla RAGBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 64.35 | 66.29 | |
| Ms-PoEBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 64.21 | 66.14 | |
| Pos2DistillBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 64.13 | 65.78 | |
| Vanilla RAGBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 63.89 | 65.43 | |
| RetRobustBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 63.49 | 65.39 | |
| RAATBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 63.12 | 65.17 | |
| Vanilla SFTBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 57.1 | 52.51 | |
| Direct GenerationBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 55.92 | 58.95 | |
| Direct GenerationBackbone=LLaMA3-8B-Instruct, Retriever=DPR2026.01 | 55.92 | 58.95 | |
| Vanilla SFTBackbone=LLaMA3-8B-Instruct, Retriever=Contriever2026.01 | 55.52 | 51.4 | |
| Vanilla SFTBackbone=Qwen3-8B, Retriever=DPR2026.01 | 54.46 | 50.17 | |
| Vanilla SFTBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 51.87 | 47.62 | |
| Naive GenerationBackbone=Qwen3-8B, Retriever=Contriever2026.01 | 45.77 | 48.16 | |
| Naive GenerationBackbone=Qwen3-8B, Retriever=DPR2026.01 | 45.77 | 48.16 |