Iterative Translation Refinement on WMT EN-FR_FR 24++ (SLI) benchmark leaderboard · SOTA2 Research

Iterative Translation Refinement on WMT EN-FR_FR 24++ (SLI)

0.6524SLI Iteration 1Qwen3-8B

Evaluation Results

Method	Links
Qwen3-8B2026.05		0.6524	0.6437	0.6374	0.6266	0.6024
Gemma4-E4B2026.05		0.6515	0.6143	0.5972	0.5992	0.5613
Gemma4-E4B2026.05		0.6478	0.6043	0.5797	0.5755	0.587
Qwen3-8B2026.05		0.6475	0.6373	0.6303	0.6243	0.6022
Qwen3-32B2026.05		0.6468	0.6353	0.6281	0.6212	0.6188
Gemma4-31B2026.05		0.6463	0.6124	0.6018	0.5787	—
Gemma4-31B2026.05		0.646	0.6102	0.6022	0.5855	0.6054
OLMo-32B2026.05		0.6439	0.6212	0.6207	0.6209	0.6201
OLMo-32B-SFT2026.05		0.6432	0.6261	0.6204	0.6209	0.6228
Qwen3-32B2026.05		0.6368	0.6147	0.6002	0.5919	0.5873
OLMo-32B-SFT2026.05		0.6254	0.6045	0.5941	0.5901	0.5868