Share your thoughts, 1 month free Claude Pro on usSee more

Commonsense Triple Validation on Benchmark ¬ATOMIC

83Valid Precision

Claude Sonnet 4

Updated 3mo ago

Evaluation Results

Method	Links
Claude Sonnet 4 2026.04		83	38	53	51	63	56	46	64	54	61	56	56	56
Llama 3.1 70B 2026.04		73	54	62	50	73	59	48	38	43	56	55	53	53
GPT-4o 2026.04		71	48	51	54	65	59	51	54	51	53	54	52	54
Llama 3.1 70B 2026.04		70	76	73	79	48	59	51	68	58	65	63	63	64
Llama 3.1 8B 2026.04		57	80	67	67	48	56	53	45	48	58	57	55	56
Llama 3.1 8B 2026.04		56	58	57	40	71	52	54	7	13	45	45	37	44