Question Answering on CovidQA (Accuracy)
67.59AccuracySFT
Evaluation results
| Method | Links | |
|---|---|---|
| SFTBackbone=Llama3.1 8B2026.03 | 67.59 | |
| SFTBackbone=Mistral 7B2026.03 | 65.97 | |
| SFTBackbone=Qwen3 1.7B2026.03 | 63.15 | |
| CoABackbone=Llama3.1 8B2026.03 | 62.14 | |
| CoABackbone=Qwen3 1.7B2026.03 | 60.87 | |
| sudoLMBackbone=Mistral 7B2026.03 | 59.04 | |
| CoABackbone=Mistral 7B2026.03 | 57.94 | |
| sudoLMBackbone=Qwen3 1.7B2026.03 | 56.41 | |
| BaseBackbone=Mistral 7B2026.03 | 52.12 | |
| PermLMBackbone=Mistral 7B2026.03 | 48.81 | |
| BaseBackbone=Qwen3 1.7B2026.03 | 48.4 | |
| BaseBackbone=Llama3.1 8B2026.03 | 45.1 | |
| PermLMBackbone=Llama3.1 8B2026.03 | 43.08 | |
| PermLMBackbone=Qwen3 1.7B2026.03 | 27.5 | |
| sudoLMBackbone=Llama3.1 8B2026.03 | 19.99 |