Carregando o concurso…
Leia o fragmento a seguir, extraído de um relatório técnico sobre sistemas de recomendação baseados em aprendizado de máquina:
"Although the model achieved a 12% improvement in click-through rate during the offline evaluation, the online A/B test revealed a statistically insignificant uplift. This discrepancy is often attributed to the fact that offline metrics fail to capture the feedback loops inherent in production environments, where user behavior is itself shaped by the recommendations they receive. Consequently, practitioners are advised to treat offline gains as a necessary but not sufficient condition for deployment, and to design online experiments that explicitly account for such dynamics."
Com base na leitura do fragmento e considerando as práticas de avaliação de sistemas tecnológicos, a afirmação que melhor sintetiza a razão pela qual os resultados offline e online podem divergir, segundo o texto, é:
Selecione uma alternativa pra criar sua conta e salvar o progresso.