Question Answering on SQuAD KRE-curated version (EM/F1)
72.6F1 ScoreAdaCAD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AdaCADCategory=Decoding-Based, Backbone=Qwen3-8B2025.10 | 72.6 | 67.4 | |
| ProbeRAGCategory=Training-Based, Backbone=Qwen3-8B2025.10 | 71.5 | 55.7 | |
| COIECDCategory=Decoding-Based, Backbone=Qwen3-8B2025.10 | 69.7 | 55.2 | |
| CANOECategory=Training-Based, Backbone=Qwen3-8B2025.10 | 69.4 | 53.4 | |
| ParamMuteCategory=Training-Based, Backbone=Qwen3-8B2025.10 | 69.2 | 52.1 | |
| ProbeRAGCategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 68.4 | 53.3 | |
| ProbeRAGCategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 68.1 | 53.6 | |
| COIECDCategory=Decoding-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 67 | 50.3 | |
| AdaCADCategory=Decoding-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 66.4 | 51.2 | |
| FaithfulRAGCategory=Prompt-Based, Backbone=Qwen3-8B2025.10 | 66.2 | 50.2 | |
| Opin(Instr)Category=Prompt-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 66 | 47.7 | |
| CANOECategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 65.4 | 49.7 | |
| Full-ContextCategory=Baseline, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 64.5 | 46 | |
| Context-DPOCategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 64.4 | 45.8 | |
| ParamMuteCategory=Training-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 64.2 | 48.5 | |
| Full-ContextCategory=Baseline, Backbone=Qwen3-8B2025.10 | 63.8 | 44.9 | |
| Opin(Instr)Category=Prompt-Based, Backbone=Qwen3-8B2025.10 | 63.8 | 46.1 | |
| Context-DPOCategory=Training-Based, Backbone=Qwen3-8B2025.10 | 63.8 | 44.9 | |
| CADCategory=Decoding-Based, Backbone=Qwen3-8B2025.10 | 63.6 | 44.5 | |
| AdaCADCategory=Decoding-Based, Backbone=Mistral-7B-v0.32025.10 | 61.6 | 43.6 | |
| FaithfulRAGCategory=Prompt-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 61.3 | 50.2 | |
| CADCategory=Decoding-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 60.3 | 41.8 | |
| FaithfulRAGCategory=Prompt-Based, Backbone=Mistral-7B-v0.32025.10 | 60.3 | 40.1 | |
| ParamMuteCategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 60.2 | 43.8 | |
| KRECategory=Prompt-Based, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 59.8 | 43.7 | |
| COIECDCategory=Decoding-Based, Backbone=Mistral-7B-v0.32025.10 | 59.2 | 39.7 | |
| Opin(Instr)Category=Prompt-Based, Backbone=Mistral-7B-v0.32025.10 | 58.1 | 37.4 | |
| CANOECategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 57.8 | 42.5 | |
| Context-DPOCategory=Training-Based, Backbone=Mistral-7B-v0.32025.10 | 56.6 | 37.6 | |
| Full-ContextCategory=Baseline, Backbone=Mistral-7B-v0.32025.10 | 56.4 | 37.5 | |
| KRECategory=Prompt-Based, Backbone=Mistral-7B-v0.32025.10 | 52.6 | 33.9 | |
| CADCategory=Decoding-Based, Backbone=Mistral-7B-v0.32025.10 | 51.4 | 32.1 | |
| KRECategory=Prompt-Based, Backbone=Qwen3-8B2025.10 | 48.6 | 29.7 | |
| No-ContextCategory=Baseline, Backbone=LLaMA-3.1-8B-Instruct2025.10 | 8.9 | 1.2 | |
| No-ContextCategory=Baseline, Backbone=Mistral-7B-v0.32025.10 | 8.1 | 1 | |
| No-ContextCategory=Baseline, Backbone=Qwen3-8B2025.10 | 6.7 | 0.4 |