Human Evaluation Translation on Sand-Glass Zh-to-En benchmark leaderboard · SOTA2 Research

Human Evaluation Translation on Sand-Glass Zh-to-En

4.97AccuracyDeepSeek-V3

Evaluation Results

Method	Links
DeepSeek-V32026.01		4.97	—	5	4.99
GPT-52026.01		4.96	—	5	4.98
Gemini-2.5-Pro2026.01		4.95	—	5	4.97
Claude-4.1-Opus2026.01		4.91	—	4.98	4.94
HOMURA_Rubric2026.01		4.91	—	4.83	4.87
HOMURA_Reason2026.01		4.85	—	4.93	4.89
GPT-52026.01		4.71	—	4.8	4.75
Gemini-2.5-Pro2026.01		4.57	—	4.57	4.57
DeepSeek-V32026.01		4.56	—	4.58	4.57
Claude-4.1-Opus2026.01		4.51	—	4.49	4.5
Gemini-2.5-Pro2026.01		4.44	—	4.9	4.67
Claude-4.1-Opus2026.01		4.42	—	4.93	4.67
GPT-52026.01		4.42	—	4.87	4.65
DeepSeek-V32026.01		4.42	—	4.87	4.65