Question Answering on Average (TVQA, SciQ, NQ-Open) (test)
74.26PrecisionFAITH
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FAITHBackbone=Llama3-8B2026.04 | 74.26 | 45.73 | |
| FAITH_sft+ppoBackbone=Llama3-8B2026.04 | 73.79 | 45.69 | |
| FAITH_sft+ppoBackbone=Mistral-7B-v0.12026.04 | 72.43 | 45.19 | |
| FAITH_sftBackbone=Llama3-8B2026.04 | 72.27 | 44.54 | |
| FAITHBackbone=Mistral-7B-v0.12026.04 | 72.22 | 45.34 | |
| FAITH_sftBackbone=Mistral-7B-v0.12026.04 | 71.85 | 44.83 | |
| InFactBackbone=Llama3-8B2026.04 | 71.33 | 44.03 | |
| Context-DPOBackbone=Llama3-8B2026.04 | 71.3 | 43.97 | |
| UAlign_ragBackbone=Llama3-8B2026.04 | 71.2 | 43.86 | |
| UAlignBackbone=Mistral-7B-v0.12026.04 | 70.82 | 41.79 | |
| UAlignBackbone=Llama3-8B2026.04 | 70.72 | 43.71 | |
| UAlign_ragBackbone=Mistral-7B-v0.12026.04 | 70.48 | 42.15 | |
| UAlign_sftBackbone=Llama3-8B2026.04 | 70.22 | 43.27 | |
| Context-DPOBackbone=Mistral-7B-v0.12026.04 | 68.89 | 43.3 | |
| InFactBackbone=Mistral-7B-v0.12026.04 | 68.59 | 43.23 | |
| ICL-CoTBackbone=Mistral-7B-v0.12026.04 | 67.69 | 39.06 | |
| UAlign_sftBackbone=Mistral-7B-v0.12026.04 | 65.61 | 41.3 | |
| ICL-CoTBackbone=Llama3-8B2026.04 | 65.45 | 40.95 | |
| SFT_ragBackbone=Mistral-7B-v0.12026.04 | 64.33 | 37.61 | |
| RL-DPOBackbone=Llama3-8B2026.04 | 64.32 | 39.11 | |
| SFTBackbone=Mistral-7B-v0.12026.04 | 63.74 | 37.08 | |
| RL-DPOBackbone=Mistral-7B-v0.12026.04 | 63.19 | 37.06 | |
| SFT_ragBackbone=Llama3-8B2026.04 | 62.45 | 38.86 | |
| SFTBackbone=Llama3-8B2026.04 | 61.46 | 38.18 |