Question Answering on CoQA (Exact Match)
71.4Exact MatchMETIS
Evaluation Results
| Method | Links | |
|---|---|---|
| METISBackbone=Gemma-2-2B2026.06 | 71.4 | |
| TIESBackbone=Gemma-2-2B2026.06 | 70.5 | |
| METISBackbone=Llama-3.1-8B2026.06 | 69.8 | |
| Pre-trainedBackbone=Llama-3.1-8B2026.06 | 69.7 | |
| Pre-trainedBackbone=Gemma-2-2B2026.06 | 69.5 | |
| DAREBackbone=Llama-3.1-8B2026.06 | 68.7 | |
| TIESBackbone=Llama-3.1-8B2026.06 | 68.7 | |
| METISBackbone=Qwen-3-4B2026.06 | 68.4 | |
| Pre-trainedBackbone=Qwen-3-4B2026.06 | 68.3 | |
| METISBackbone=Llama-3.2-3B2026.06 | 67 | |
| TIESBackbone=Llama-3.2-3B2026.06 | 66.2 | |
| TIESBackbone=Qwen-3-4B2026.06 | 66.1 | |
| Pre-trainedBackbone=Llama-3.2-3B2026.06 | 65.9 | |
| ConsensusTABackbone=Gemma-2-2B2026.06 | 64.5 | |
| DAREBackbone=Qwen-3-4B2026.06 | 63.9 | |
| ConsensusTABackbone=Llama-3.2-3B2026.06 | 58.3 | |
| ConsensusTABackbone=Llama-3.1-8B2026.06 | 57.7 | |
| Task ArithmeticBackbone=Llama-3.1-8B2026.06 | 57.3 | |
| DAREBackbone=Gemma-2-2B2026.06 | 56.1 | |
| DAREBackbone=Llama-3.2-3B2026.06 | 55.2 | |
| Task ArithmeticBackbone=Llama-3.2-3B2026.06 | 54.8 | |
| ConsensusTABackbone=Qwen-3-4B2026.06 | 52.1 | |
| Task ArithmeticBackbone=Qwen-3-4B2026.06 | 47.3 | |
| Task ArithmeticBackbone=Gemma-2-2B2026.06 | 42 |