Question Answering on WebQ-QA Out-of-Domain (test)
67.99PrecisionFAITH
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FAITHBackbone=Llama3-8B2026.04 | 67.99 | 34.03 | |
| FAITH_sft+ppoBackbone=Llama3-8B2026.04 | 67.31 | 33.75 | |
| UAlign_ragBackbone=Llama3-8B2026.04 | 67.01 | 33.21 | |
| FAITH_sftBackbone=Llama3-8B2026.04 | 66.94 | 33.1 | |
| SFT_ragBackbone=Llama3-8B2026.04 | 66.93 | 31.85 | |
| InFactBackbone=Llama3-8B2026.04 | 66.92 | 33.44 | |
| UAlignBackbone=Llama3-8B2026.04 | 66.88 | 33.01 | |
| Context-DPOBackbone=Llama3-8B2026.04 | 66.75 | 33.12 | |
| SFTBackbone=Llama3-8B2026.04 | 66.46 | 31.18 | |
| UAlign_sftBackbone=Llama3-8B2026.04 | 66.12 | 32.58 | |
| RL-DPOBackbone=Llama3-8B2026.04 | 65.99 | 32.41 | |
| ICL-CoTBackbone=Llama3-8B2026.04 | 65.97 | 30.85 | |
| DTABackbone=Llama3-8B2026.04 | 61.24 | 30.71 | |
| FAITHBackbone=Mistral-7B-v0.12026.04 | 58.04 | 40.43 | |
| Context-DPOBackbone=Mistral-7B-v0.12026.04 | 57.32 | 37.99 | |
| InFactBackbone=Mistral-7B-v0.12026.04 | 56.98 | 37.73 | |
| UAlign_ragBackbone=Mistral-7B-v0.12026.04 | 56.97 | 37.79 | |
| DTABackbone=Mistral-7B-v0.12026.04 | 56.53 | 23.44 | |
| UAlignBackbone=Mistral-7B-v0.12026.04 | 56.47 | 37.02 | |
| FAITH_sft+ppoBackbone=Mistral-7B-v0.12026.04 | 55.48 | 38.65 | |
| UAlign_sftBackbone=Mistral-7B-v0.12026.04 | 55.34 | 36.87 | |
| ICL-CoTBackbone=Mistral-7B-v0.12026.04 | 53.43 | 35.76 | |
| SFT_ragBackbone=Mistral-7B-v0.12026.04 | 53.14 | 34.82 | |
| SFTBackbone=Mistral-7B-v0.12026.04 | 52.24 | 34.33 | |
| RL-DPOBackbone=Mistral-7B-v0.12026.04 | 52.01 | 33.87 | |
| FAITH_sftBackbone=Mistral-7B-v0.12026.04 | 51.99 | 41.77 |