Question Answering on TVQA In-Domain (test)
87.2PrecisionFAITH
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FAITHBackbone=Mistral-7B-v0.12026.04 | 87.2 | 60.72 | |
| FAITH_sft+ppoBackbone=Mistral-7B-v0.12026.04 | 87 | 60.58 | |
| FAITH_sftBackbone=Mistral-7B-v0.12026.04 | 86.51 | 60.24 | |
| Context-DPOBackbone=Mistral-7B-v0.12026.04 | 84.32 | 59.51 | |
| FAITHBackbone=Llama3-8B2026.04 | 84.19 | 60.69 | |
| InFactBackbone=Mistral-7B-v0.12026.04 | 83.44 | 59.32 | |
| FAITH_sft+ppoBackbone=Llama3-8B2026.04 | 82.95 | 59.8 | |
| UAlign_ragBackbone=Mistral-7B-v0.12026.04 | 82.74 | 59.55 | |
| UAlignBackbone=Mistral-7B-v0.12026.04 | 82.1 | 59.05 | |
| FAITH_sftBackbone=Llama3-8B2026.04 | 81.24 | 58.77 | |
| UAlign_sftBackbone=Mistral-7B-v0.12026.04 | 81.07 | 56.47 | |
| InFactBackbone=Llama3-8B2026.04 | 80.11 | 57.76 | |
| Context-DPOBackbone=Llama3-8B2026.04 | 79.76 | 57.26 | |
| UAlign_ragBackbone=Llama3-8B2026.04 | 79.35 | 57.24 | |
| UAlignBackbone=Llama3-8B2026.04 | 79.14 | 57.04 | |
| UAlign_sftBackbone=Llama3-8B2026.04 | 78.76 | 56.68 | |
| ICL-CoTBackbone=Mistral-7B-v0.12026.04 | 76.73 | 54.78 | |
| SFT_ragBackbone=Mistral-7B-v0.12026.04 | 75.87 | 55.09 | |
| SFTBackbone=Mistral-7B-v0.12026.04 | 74.57 | 54.77 | |
| RL-DPOBackbone=Mistral-7B-v0.12026.04 | 72.2 | 52.98 | |
| RL-DPOBackbone=Llama3-8B2026.04 | 72.08 | 53.96 | |
| SFT_ragBackbone=Llama3-8B2026.04 | 71.49 | 53.04 | |
| SFTBackbone=Llama3-8B2026.04 | 70.8 | 52.57 | |
| ICL-CoTBackbone=Llama3-8B2026.04 | 66.68 | 53.37 | |
| DTABackbone=Llama3-8B2026.04 | 43.99 | 31.73 | |
| DTABackbone=Mistral-7B-v0.12026.04 | 41.33 | 28.78 |