Faithful Calibration on PQA
89cMFG*Qwen3-1.7B + RLMF + Rewr.
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-1.7B + RLMF + Rewr.Variant=RLMF + Rewriting2026.06 | 89 | |
| Llama3.1-8B-Ins + RLMFVariant=RLMF2026.06 | 85 | |
| Qwen3-8B + RLMFVariant=RLMF2026.06 | 85 | |
| Qwen3-1.7B + RLMFVariant=RLMF2026.06 | 83 | |
| Llama3.1-8B-Ins + RLVariant=RL2026.06 | 82 | |
| Qwen3-8B + RLMF + Rewr.Variant=RLMF + Rewriting2026.06 | 82 | |
| Llama3.1-8B-Ins + RLMF + Rewr.Variant=RLMF + Rewriting2026.06 | 81 | |
| Qwen3-4B + RLMFVariant=RLMF2026.06 | 81 | |
| Qwen3-4B + RLVariant=RL2026.06 | 80 | |
| Qwen3-4B + RLMF + Rewr.Variant=RLMF + Rewriting2026.06 | 80 | |
| Qwen3-1.7B + RLVariant=RL2026.06 | 78 | |
| Qwen3-8B + RLVariant=RL2026.06 | 75 | |
| Llama3.1-8B-Ins + FUTVariant=FUT2026.06 | 69 | |
| Llama3.1-8B-Ins + MetaFaithVariant=MetaFaith2026.06 | 68 | |
| Gemini-3.1-ProVariant=Base2026.06 | 62 | |
| Llama3.1-8B-InsVariant=Base2026.06 | 60 | |
| Gemini-3-FlashVariant=Base2026.06 | 59 | |
| Qwen3-8B + FUTVariant=FUT2026.06 | 57 | |
| Qwen3-1.7BVariant=Base2026.06 | 57 | |
| Qwen3-1.7B + FUTVariant=FUT2026.06 | 55 | |
| Qwen3-8BVariant=Base2026.06 | 53 | |
| Qwen3-8B + MetaFaithVariant=MetaFaith2026.06 | 53 | |
| Qwen3-4BVariant=Base2026.06 | 53 | |
| Qwen3-1.7B + MetaFaithVariant=MetaFaith2026.06 | 50 | |
| GPT-5Variant=Base2026.06 | 50 | |
| Qwen3-4B + MetaFaithVariant=MetaFaith2026.06 | 48 | |
| Qwen3-4B + FUTVariant=FUT2026.06 | 47 |