Truthfulness Evaluation on TruthfulQA (Normalized Accuracy)
58.76Normalized AccuracyIPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IPOBase model=Qwen-2 instruct 7B2025.05 | 58.76 | — | |
| SamPOBase model=Qwen-2 instruct 7B2025.05 | 58.44 | — | |
| SGDPOBase model=Qwen-2 instruct 7B2025.05 | 58.06 | — | |
| NCABase model=Qwen-2 instruct 7B2025.05 | 57.82 | — | |
| BCOBase model=Qwen-2 instruct 7B2025.05 | 57.76 | — | |
| DPOBase model=Qwen-2 instruct 7B2025.05 | 57.74 | — | |
| TDPOBase model=Qwen-2 instruct 7B2025.05 | 57.63 | — | |
| SFTBase model=Qwen-2 instruct 7B2025.05 | 57.34 | — | |
| Llama 3-8B E8T2Shots=0-shot2024.12 | 44.22 | — | |
| Llama 3-8BShots=0-shot2024.12 | 44.01 | — | |
| HATified-SFTShots=6-shot2026.03 | — | 17.6 | |
| Llama-InstructShots=6-shot2026.03 | — | 19.1 |