Grammatical Error Correction on Falko-Merlin
64.5PrecisionLlama-3.1-8B-Instruct
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama-3.1-8B-InstructDecoding=edit-level majority voting2026.05 | 64.5 | 32.2 | 53.7 | |
| Qwen3-8BDecoding=MBR2026.05 | 62.5 | 38.2 | 55.5 | |
| Qwen3-8BDecoding=Greedy2026.05 | 61.5 | 39.9 | 55.5 | |
| gemma-3-12b-itDecoding=MBR2026.05 | 58.3 | 52.6 | 57 | |
| gemma-3-12b-itDecoding=edit-level majority voting2026.05 | 57.6 | 39.6 | 52.8 | |
| gemma-3-12b-itDecoding=Greedy2026.05 | 57.2 | 53.9 | 56.5 | |
| Qwen3-8BDecoding=edit-level majority voting2026.05 | 57.1 | 30.7 | 48.7 | |
| Llama-3.1-8B-InstructDecoding=Greedy2026.05 | 55.2 | 51.3 | 54.4 | |
| Llama-3.1-8B-InstructDecoding=MBR2026.05 | 53.6 | 50 | 52.8 |