Question Answering on Average of 5 Datasets
78.9Average ScoreGUARANTRAG
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GUARANTRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 78.9 | — | — | — | — | |
| GUARANTRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 77.1 | — | — | — | — | |
| GUARANTRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 76.6 | — | — | — | — | |
| P-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 75.8 | — | — | — | — | |
| P-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 74.4 | — | — | — | — | |
| GUARANTRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 74.2 | — | — | — | — | |
| SOLARBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 74.1 | — | — | — | — | |
| P-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 73.7 | — | — | — | — | |
| DRAGINBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 73.6 | — | — | — | — | |
| FLAREBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 73.1 | — | — | — | — | |
| SOLARBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 72.5 | — | — | — | — | |
| RQ-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 72.4 | — | — | — | — | |
| DRAGINBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 72 | — | — | — | — | |
| SOLARBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 71.6 | — | — | — | — | |
| FLAREBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 71.6 | — | — | — | — | |
| GUARANTRAG-7BRetriever=BM252026.04 | 71.6 | — | — | — | — | |
| P-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 71.3 | — | — | — | — | |
| DRAGINBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 71.2 | — | — | — | — | |
| FLAREBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 70.9 | — | — | — | — | |
| RQ-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 70.8 | — | — | — | — | |
| SelfRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 70.6 | — | — | — | — | |
| GUARANTRAGModel Variant=w/o DPO Training, Retriever=BM252026.04 | 70.5 | — | — | — | — | |
| RQ-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 70.1 | — | — | — | — | |
| SOLARBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 69.4 | — | — | — | — | |
| DA-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 69.4 | — | — | — | — | |
| SelfRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 69.1 | — | — | — | — | |
| DRAGINBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 69.1 | — | — | — | — | |
| GUARANTRAGModel Variant=w/o Relevance Control, Retriever=BM252026.04 | 69.1 | — | — | — | — | |
| GUARANTRAGModel Variant=w/o Segment-Level Fusion, Retriever=BM252026.04 | 68.9 | — | — | — | — | |
| FLAREBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 68.7 | — | — | — | — | |
| SelfRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 68.5 | — | — | — | — | |
| DA-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 68.2 | — | — | — | — | |
| RQ-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 67.9 | — | — | — | — | |
| DA-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 67.3 | — | — | — | — | |
| GUARANTRAG-4BRetriever=BM252026.04 | 67.3 | — | — | — | — | |
| GUARANTRAGModel Variant=w/o Length Control, Retriever=BM252026.04 | 66.8 | — | — | — | — | |
| SelfRAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 66.2 | — | — | — | — | |
| RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=RetroMAE2026.04 | 65.7 | — | — | — | — | |
| DA-RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 65.1 | — | — | — | — | |
| RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=SPLADE2026.04 | 64.3 | — | — | — | — | |
| RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=HyDE2026.04 | 64 | — | — | — | — | |
| RAGBase Model=Qwen3-14B-T, Thinking Mode=true, Retriever=BM252026.04 | 62.1 | — | — | — | — | |
| Qwen3-14B-TThinking Mode=true, Retriever=BM252026.04 | 59.5 | — | — | — | — | |
| Qwen3-14B-TThinking Mode=true, Retriever=SPLADE2026.04 | 59.5 | — | — | — | — | |
| Qwen3-14B-TThinking Mode=true, Retriever=RetroMAE2026.04 | 59.5 | — | — | — | — | |
| Qwen3-14B-TThinking Mode=true, Retriever=HyDE2026.04 | 59.5 | — | — | — | — | |
| CompActReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 30.3 | 39.4 | 0.2 | 96.3 | |
| EXITReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 30.1 | 38.4 | 0.7 | 56.3 | |
| LongLLMLinguaReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 31.3 | 40 | 0.7 | 60.8 | |
| LooCompReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 32.4 | 41.4 | 6.3 | 90.3 | |
| ProvenceReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 31 | 39.8 | 5.2 | 84.5 | |
| RECOMP-absReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 29 | 38 | 0.8 | 98.2 | |
| RECOMP-extReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 25.8 | 33.6 | 27.7 | 88.1 | |
| RefinerReader=Llama-3.1-8B Instruct, top-k chunks=202026.03 | — | 29.3 | 37.8 | 0.2 | 94.3 |