Machine Translation on lexical choice (val+test) benchmark leaderboard · SOTA2 Research

Machine Translation on lexical choice (val+test)

54.15BLEUgpt-4o

Evaluation Results

Method	Links
gpt-4o2025.10		54.15	69.41	0.9208	0.8797	0.31
gpt-42025.10		53.99	68.79	0.9179	0.8713	2.58
gpt-4o2025.10		53.84	69.67	0.9223	0.8851	—
gpt-42025.10		51.41	66.63	0.9128	0.8643	—
gpt-4-turbo2025.10		49.56	65.67	0.912	0.8683	—
gpt-3.5-turbo2025.10		49.42	65.69	0.9102	0.856	2.11
gpt-4-turbo2025.10		49.25	64.84	0.8911	0.8557	-0.31
gpt-3.5-turbo2025.10		47.31	64.88	0.9061	0.8619	—
Llama3.32025.10		46.58	63.37	0.9005	0.8449	0.15
Llama3.32025.10		46.43	62.47	0.9023	0.8479	—
Phi-42025.10		44.35	61.37	0.9	0.8355	1.77
Phi-42025.10		42.58	59.63	0.8936	0.8334	—
DeepSeek-R1 32B2025.10		40.94	58.26	0.8816	0.814	2.07
DeepSeek-R1 32B2025.10		38.87	56.25	0.8827	0.8194	—
DeepSeek-R1 14B2025.10		38.43	55.13	0.8732	0.8031	—
Llama3.12025.10		35.08	52.1	0.8717	0.7925	—
DeepSeek-R1 14B2025.10		33.33	51.14	0.8506	0.7701	-5.1
DeepSeek-R1 8B2025.10		30.89	48.33	0.8558	0.7632	—
DeepSeek-R1 8B2025.10		30.26	48.05	0.8331	0.7517	-0.63
Llama3.12025.10		29.37	48.57	0.8573	0.7664	-5.71
Mistral2025.10		27.79	45	0.849	0.7537	1.17
Mistral2025.10		26.62	44.89	0.8444	0.7436	—
Llama3.22025.10		25.19	42.47	0.8311	0.7297	—
Llama3.22025.10		22.38	43.1	0.819	0.7041	-2.81