Search
Meta Executive Denies Boosting Llama 4's Benchmark Scores Artificially
2025-04-08 /
Read about 0 minute
Author：小编
In an article published on Monday, Ahmad Al-Dahle, Vice President of AI Generation at Meta, firmly denied allegations that the company had manipulated its new AI models, Llama 4 Maverick and Llama 4 Scout, to excel in specific benchmark tests and conceal their potential weaknesses. He explicitly stated that Meta had not specifically trained these models on the 'test set,' asserting that the claim was 'simply not true.'
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
