Question Answering on WebQA
99.02ASRVanilla
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VanillaPoisoning Method=SB2026.05 | 99.02 | — | — | — | — | 42.72 | |
| FreqFedPoison Method=Addsent2026.02 | 98.92 | — | — | — | 45.77 | — | |
| CUBEPoison Method=Addsent2026.02 | 98.38 | — | — | — | 45.47 | — | |
| FreqFedPoison Method=Badnets2026.02 | 98.23 | — | — | — | 43.95 | — | |
| NonePoison Method=Badnets2026.02 | 98.08 | — | — | — | 46.36 | — | |
| CUBEPoison Method=Badnets2026.02 | 97.93 | — | — | — | 43.95 | — | |
| KrumPoison Method=Addsent2026.02 | 97.79 | — | — | — | 45.42 | — | |
| CUBEPoison Method=StyleBkd2026.02 | 97.5 | — | — | — | 42.62 | — | |
| MedianPoison Method=Badnets2026.02 | 97.29 | — | — | — | 46.7 | — | |
| TrimmedPoison Method=Badnets2026.02 | 97.05 | — | — | — | 46.65 | — | |
| KrumPoison Method=Badnets2026.02 | 96.75 | — | — | — | 45.32 | — | |
| ONIONPoison Method=CBA2026.02 | 96.6 | — | — | — | 9.74 | — | |
| ONIONPoison Method=Addsent2026.02 | 96.16 | — | — | — | 9.65 | — | |
| MedianPoison Method=Addsent2026.02 | 95.52 | — | — | — | 46.46 | — | |
| NonePoison Method=Addsent2026.02 | 95.42 | — | — | — | 45.72 | — | |
| ONIONPoison Method=Badnets2026.02 | 95.32 | — | — | — | 9.79 | — | |
| TrimmedPoison Method=Addsent2026.02 | 95.13 | — | — | — | 46.56 | — | |
| KrumPoison Method=StyleBkd2026.02 | 94.33 | — | — | — | 44.19 | — | |
| TrimmedPoison Method=StyleBkd2026.02 | 94 | — | — | — | 44.14 | — | |
| ONIONPoison Method=StyleBkd2026.02 | 93.67 | — | — | — | 8.46 | — | |
| MedianPoison Method=StyleBkd2026.02 | 93.33 | — | — | — | 44.39 | — | |
| FreqFedPoison Method=StyleBkd2026.02 | 93.33 | — | — | — | 44 | — | |
| ONIONPoisoning Method=SB2026.05 | 92.62 | — | — | — | — | 18.16 | |
| NonePoison Method=StyleBkd2026.02 | 92.33 | — | — | — | 41.63 | — | |
| FoundationFLPoison Method=StyleBkd2026.02 | 92 | — | — | — | 42.32 | — | |
| VanillaPoisoning Method=CBA2026.05 | 91.38 | — | — | — | — | 42.32 | |
| FoundationFLPoison Method=Badnets2026.02 | 89.71 | — | — | — | 44.49 | — | |
| VanillaPoisoning Method=BN2026.05 | 84.55 | — | — | — | — | 39.37 | |
| GraCeFulPoison Method=Badnets2026.02 | 79.72 | — | — | — | 43.95 | — | |
| FoundationFLPoison Method=Addsent2026.02 | 56 | — | — | — | 45.03 | — | |
| VanillaPoisoning Method=AS2026.05 | 49.75 | — | — | — | — | 41.29 | |
| FoundationFLPoison Method=CBA2026.02 | 46.21 | — | — | — | 42.86 | — | |
| GraCeFulPoison Method=Addsent2026.02 | 31.89 | — | — | — | 43.41 | — | |
| GraCeFulPoison Method=StyleBkd2026.02 | 30.33 | — | — | — | 45.23 | — | |
| GraCeFulPoison Method=CBA2026.02 | 30.29 | — | — | — | 42.67 | — | |
| NonePoison Method=CBA2026.02 | 21.46 | — | — | — | 45.08 | — | |
| TrimmedPoison Method=CBA2026.02 | 19.1 | — | — | — | 45.08 | — | |
| CUBEPoison Method=CBA2026.02 | 16.73 | — | — | — | 45.23 | — | |
| KrumPoison Method=CBA2026.02 | 13.24 | — | — | — | 44.29 | — | |
| MedianPoison Method=CBA2026.02 | 9.84 | — | — | — | 44.78 | — | |
| FreqFedPoison Method=CBA2026.02 | 6.25 | — | — | — | 43.85 | — | |
| ONIONPoisoning Method=BN2026.05 | 5.91 | — | — | — | — | 25.84 | |
| ONIONPoisoning Method=CBA2026.05 | 1.48 | — | — | — | — | 29.38 | |
| ONIONPoisoning Method=AS2026.05 | 1.08 | — | — | — | — | 26.97 | |
| CleanGenPoisoning Method=CBA2026.05 | 0.3 | — | — | — | — | 29.38 | |
| CleanGenPoisoning Method=BN2026.05 | 0.2 | — | — | — | — | 27.95 | |
| CleanGenPoisoning Method=SB2026.05 | 0.2 | — | — | — | — | 22.79 | |
| CleanGenPoisoning Method=AS2026.05 | 0.1 | — | — | — | — | 27.76 | |
| ProtegoFedPoison Method=Badnets2026.02 | 0 | — | — | — | 45.37 | — | |
| ProtegoFedPoison Method=Addsent2026.02 | 0 | — | — | — | 45.28 | — | |
| ProtegoFedPoison Method=CBA2026.02 | 0 | — | — | — | 45.28 | — | |
| ProtegoFedPoison Method=StyleBkd2026.02 | 0 | — | — | — | 45.62 | — | |
| CUBEPoisoning Method=BN2026.05 | 0 | — | — | — | — | 38.73 | |
| GraCeFulPoisoning Method=BN2026.05 | 0 | — | — | — | — | 39.37 | |
| GradSentryPoisoning Method=BN2026.05 | 0 | — | — | — | — | 39.67 | |
| CUBEPoisoning Method=AS2026.05 | 0 | — | — | — | — | 38.04 | |
| GraCeFulPoisoning Method=AS2026.05 | 0 | — | — | — | — | 38.78 | |
| GradSentryPoisoning Method=AS2026.05 | 0 | — | — | — | — | 39.62 | |
| CUBEPoisoning Method=CBA2026.05 | 0 | — | — | — | — | 38.19 | |
| GraCeFulPoisoning Method=CBA2026.05 | 0 | — | — | — | — | 41.09 | |
| GradSentryPoisoning Method=CBA2026.05 | 0 | — | — | — | — | 42.57 | |
| CUBEPoisoning Method=SB2026.05 | 0 | — | — | — | — | 37.8 | |
| GraCeFulPoisoning Method=SB2026.05 | 0 | — | — | — | — | 39.52 | |
| GradSentryPoisoning Method=SB2026.05 | 0 | — | — | — | — | 41.39 | |
| CoTGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=32026.01 | — | 25.84 | 39.99 | — | — | — | |
| CoTGenerator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=52026.01 | — | 26.18 | 40.54 | — | — | — | |
| DTRGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=3, Uncertainty threshold (u)=0.0012026.01 | — | 26.18 | 40.65 | 91.8 | — | — | |
| DTRGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=3, Uncertainty threshold (u)=0.0052026.01 | — | 26.97 | 41.41 | 84.5 | — | — | |
| DTRGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=3, Uncertainty threshold (u)=0.012026.01 | — | 27.36 | 41.57 | 81.1 | — | — | |
| DTRGenerator=Qwen2.5-7B-Instruct, Threshold u=0.0012026.01 | — | 26.28 | 40.65 | 91.8 | — | — | |
| DTRGenerator=Qwen2.5-72B-Instruct, Threshold u=0.0012026.01 | — | 24.56 | 42.26 | 82.7 | — | — | |
| DTRGenerator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=5, Uncertainty threshold (u)=0.0012026.01 | — | 26.48 | 41.23 | 91.8 | — | — | |
| DTRGenerator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=5, Uncertainty threshold (u)=0.0052026.01 | — | 27.21 | 41.9 | 84.5 | — | — | |
| DTRGenerator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=5, Uncertainty threshold (u)=0.012026.01 | — | 27.56 | 42.1 | 81.1 | — | — | |
| GLM-4-VoiceBackbone=GLM-4-Voice, Text EE=-, Speech EE=-2026.03 | — | — | — | — | — | 52.2 | |
| GLM-4-VoiceBackbone=GLM-4-Voice, Text EE=-, Speech EE=Fixed (38)2026.03 | — | — | — | — | — | 51.4 | |
| GLM-4-VoiceBackbone=GLM-4-Voice, Text EE=-, Speech EE=Conf (36)2026.03 | — | — | — | — | — | 51 | |
| HyDEGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=32026.01 | — | 25.69 | 39.4 | — | — | — | |
| HyDEGenerator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=52026.01 | — | 27.07 | 41 | — | — | — | |
| LLM JudgeGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=3, Judge Model=7B2026.01 | — | 22.59 | 35.03 | 0 | — | — | |
| LLM JudgeGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=3, Judge Model=72B2026.01 | — | 24.16 | 36.93 | 26.9 | — | — | |
| LLM Judge (72B)Generator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=52026.01 | — | 24.11 | 36.73 | 26.9 | — | — | |
| LLM Judge (7B)Generator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=52026.01 | — | 22.59 | 35.03 | 0 | — | — | |
| No RetrievalGenerator=Qwen2.5-7B-Instruct, Retriever=None, Top-k retrievals=02026.01 | — | 22.59 | 35.03 | — | — | — | |
| No RetrievalGenerator=Qwen2.5-7B-Instruct, Retriever=None2026.01 | — | 22.59 | 35.03 | — | — | — | |
| Q2DGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=32026.01 | — | 25.39 | 39.62 | — | — | — | |
| Q2DGenerator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=52026.01 | — | 26.03 | 40.18 | — | — | — | |
| SPAR-KBackbone=Step-Audio-2, Text EE=-, Speech EE=Even (22)2026.03 | — | — | — | — | — | 55.4 | |
| SPAR-KBackbone=Step-Audio-2, Text EE=-, Speech EE=Odd (22)2026.03 | — | — | — | — | — | 53.3 | |
| SPAR-KBackbone=Step-Audio-2, Text EE=-, Speech EE=Triple (22)2026.03 | — | — | — | — | — | 56.5 | |
| SPAR-KBackbone=GLM-4-Voice, Text EE=-, Speech EE=Even (36)2026.03 | — | — | — | — | — | 50.6 | |
| SPAR-KBackbone=GLM-4-Voice, Text EE=-, Speech EE=Odd (36)2026.03 | — | — | — | — | — | 50.3 | |
| SPAR-KBackbone=GLM-4-Voice, Text EE=-, Speech EE=Triple (37)2026.03 | — | — | — | — | — | 51.9 | |
| SPAR-KBackbone=GLM-4-Voice, Text EE=-, Speech EE=Even (33)2026.03 | — | — | — | — | — | 48.5 | |
| SPAR-KBackbone=GLM-4-Voice, Text EE=-, Speech EE=Even (35)2026.03 | — | — | — | — | — | 49.5 | |
| SPAR-KBackbone=GLM-4-Voice, Text EE=Even (36), Speech EE=-2026.03 | — | — | — | — | — | 21.7 | |
| SPAR-KBackbone=GLM-4-Voice, Text EE=Conf (36), Speech EE=Even(36)2026.03 | — | — | — | — | — | 48.3 | |
| Standard RAGGenerator=Qwen2.5-7B-Instruct, Retriever=e5, Top-k retrievals=32026.01 | — | 23.72 | 38.43 | — | — | — | |
| Standard RAGGenerator=Qwen2.5-7B-Instruct2026.01 | — | 24.7 | 38.27 | — | — | — | |
| Standard RAGGenerator=Qwen2.5-7B-Instruct, Retriever=bge, Retrieval depth (top-k)=52026.01 | — | 25.15 | 39.29 | — | — | — |