Question Answering on TruthfulQA (MC1, MC2, Bleu, Bleurt)
36.5MC1 ScoreITI
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ITIBackbone=Llama3-8b-inst., Method=ITI2025.08 | 36.5 | 54.29 | 43.22 | 66.3 | |
| MSRSBackbone=Llama3-8b-inst., Method=MSRS2025.08 | 34.91 | 56.32 | 52.32 | 66.75 | |
| CS-ReFTBackbone=Llama3-8b-inst., Method=CS-ReFT2025.08 | 34.82 | 53.23 | 52.18 | 63.26 | |
| MSRSBackbone=Qwen2-7b-inst., Method=MSRS2025.08 | 34.72 | 53.27 | 53.1 | 74.9 | |
| MSRSBackbone=Mistral-7b-v0.3, Method=MSRS2025.08 | 31.32 | 52.62 | 50.61 | 71.39 | |
| ReFTBackbone=Mistral-7b-v0.3, Method=ReFT2025.08 | 30.07 | 49.69 | 49.39 | 66.01 | |
| ReFTBackbone=Qwen2-7b-inst., Method=ReFT2025.08 | 29.83 | 48.69 | 52.57 | 71.15 | |
| ReFTBackbone=Llama3-8b-inst., Method=ReFT2025.08 | 29.58 | 49.51 | 52.08 | 64.06 | |
| CS-ReFTBackbone=Mistral-7b-v0.3, Method=CS-ReFT2025.08 | 29.45 | 51.12 | 49.95 | 69.2 | |
| MAT-STEERBackbone=Llama3-8b-inst., Method=MAT-STEER2025.08 | 29.29 | 49.67 | 43.81 | 55.31 | |
| CS-ReFTBackbone=Qwen2-7b-inst., Method=CS-ReFT2025.08 | 29.15 | 48.92 | 52.85 | 72.25 | |
| CAABackbone=Mistral-7b-v0.3, Method=CAA2025.08 | 28.77 | 52.05 | 53.85 | 62.27 | |
| CAABackbone=Qwen2-7b-inst., Method=CAA2025.08 | 28.44 | 47.25 | 48.35 | 58.97 | |
| CAABackbone=Llama3-8b-inst., Method=CAA2025.08 | 28.41 | 47.55 | 45.42 | 61.54 | |
| ICLBackbone=Llama3-8b-inst., Method=ICL2025.08 | 28.37 | 46.21 | 48.35 | 60.44 | |
| Llama3-8b-inst.Backbone=Llama3-8b-inst., Method=Base2025.08 | 27.47 | 45.63 | 46.89 | 57.88 | |
| ICLBackbone=Qwen2-7b-inst., Method=ICL2025.08 | 26.84 | 48.33 | 49.79 | 66.63 | |
| Qwen2-7b-inst.Backbone=Qwen2-7b-inst., Method=Base2025.08 | 26.38 | 45.41 | 49.63 | 65.28 | |
| MAT-STEERBackbone=Mistral-7b-v0.3, Method=MAT-STEER2025.08 | 25.45 | 48.38 | 49.46 | 62.62 | |
| MAT-STEERBackbone=Qwen2-7b-inst., Method=MAT-STEER2025.08 | 23.08 | 49.51 | 51.72 | 72.53 | |
| ICLBackbone=Mistral-7b-v0.3, Method=ICL2025.08 | 21.92 | 49.23 | 44.11 | 57.64 | |
| Mistral-7b-v0.3Backbone=Mistral-7b-v0.3, Method=Base2025.08 | 18.83 | 36.54 | 41.56 | 54.52 |