Faithfulness Evaluation on FaithEval
74.9F1 ScoreProbeRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ProbeRAGCategory=Training-Based, Backbone=Qwen3-8B2025.10 | 74.9 | 61.6 | |
| ProbeRAGCategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 74.9 | 62.9 | |
| ProbeRAGCategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 74.4 | 64.4 | |
| CANOECategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 71.6 | 56.3 | |
| CANOEArchitecture=LLaMA-2-7B-Chat-HF2025.10 | 70.6 | 52.3 | |
| CANOECategory=Training-Based, Backbone=Qwen3-8B2025.10 | 70.3 | 60.2 | |
| FaithfulRAGCategory=Prompt-Based, Backbone=Qwen3-8B2025.10 | 70.2 | 58.3 | |
| CADCategory=Decoding-Based, Backbone=Mistral-7B-v0.32025.10 | 68.9 | 33.3 | |
| Full-ContextCategory=Baseline, Backbone=Mistral-7B-v0.32025.10 | 68.8 | 37.7 | |
| ParamMuteCategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 68.5 | 56.2 | |
| AdaCADCategory=Decoding-Based, Backbone=Mistral-7B-v0.32025.10 | 68.4 | 34.2 | |
| ProbeRAGArchitecture=LLaMA-2-7B-Chat-HF2025.10 | 68.3 | 54.4 | |
| CANOEArchitecture=Qwen2.5-7B-Instruct2025.10 | 68.1 | 53.9 | |
| ParamMuteCategory=Training-Based, Backbone=Qwen3-8B2025.10 | 67.8 | 58.2 | |
| AdaCADCategory=Decoding-Based, Backbone=Qwen3-8B2025.10 | 67.3 | 58.2 | |
| Context-DPOCategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 67.2 | 53.7 | |
| ParamMuteCategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 67.1 | 45.2 | |
| Full-ContextCategory=Baseline, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 66.9 | 53.1 | |
| COIECDCategory=Decoding-Based, Backbone=Qwen3-8B2025.10 | 66.6 | 56.4 | |
| FaithfulRAGCategory=Prompt-Based, Backbone=Mistral-7B-v0.32025.10 | 66.4 | 40.3 | |
| Context-DPOArchitecture=Qwen2.5-7B-Instruct2025.10 | 65.1 | 50.2 | |
| Context-DPOCategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 64.9 | 31.8 | |
| KRECategory=Prompt-Based, Backbone=Mistral-7B-v0.32025.10 | 64.8 | 36.5 | |
| COIECDCategory=Decoding-Based, Backbone=Mistral-7B-v0.32025.10 | 64.4 | 29.5 | |
| CANOECategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 64.1 | 44.9 | |
| ProbeRAGArchitecture=Qwen2.5-7B-Instruct2025.10 | 63.5 | 48.9 | |
| Context-DPOArchitecture=LLaMA-2-7B-Chat-HF2025.10 | 63.2 | 50.7 | |
| FaithfulRAGCategory=Prompt-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 62.4 | 49.1 | |
| AdaCADCategory=Decoding-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 61.4 | 47.7 | |
| CADCategory=Decoding-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 59.4 | 42.7 | |
| KRECategory=Prompt-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 59.1 | 12.1 | |
| KRECategory=Prompt-Based, Backbone=Qwen3-8B2025.10 | 58.1 | 12.3 | |
| CADCategory=Decoding-Based, Backbone=Qwen3-8B2025.10 | 57 | 28.7 | |
| COIECDCategory=Decoding-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 56.1 | 41.3 | |
| Full-ContextCategory=Baseline, Backbone=Qwen3-8B2025.10 | 55.5 | 23.8 | |
| Context-DPOCategory=Training-Based, Backbone=Qwen3-8B2025.10 | 55.2 | 24 | |
| Opin(Instr)Category=Prompt-Based, Backbone=Mistral-7B-v0.32025.10 | 35.7 | 14.1 | |
| Opin(Instr)Category=Prompt-Based, Backbone=Qwen3-8B2025.10 | 35 | 13.9 | |
| Opin(Instr)Category=Prompt-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 34.9 | 15.1 | |
| No-ContextCategory=Baseline, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 27.7 | 6 | |
| No-ContextCategory=Baseline, Backbone=Mistral-7B-v0.32025.10 | 26.2 | 4.4 | |
| No-ContextCategory=Baseline, Backbone=Qwen3-8B2025.10 | 22.8 | 4.1 |