A principal inovação dos modelos transformers é o uso do mecanismo de autoatenção (self-attention). Isso permite que o modelo: Considere o contexto completo de uma sequência de entrada (por exemplo, uma frase inteira), atribuindo pesos diferentes a cada palavra, dependendo de sua relevância.Capturar relacionamentos entre palavras distantes no texto, o que é crucial para o entendimento da linguagem natural.

C. Transformer-based language models use self-attention mechanisms to capture contextual relationships. Explanation:Self-attention: This is a key feature of transformers that allows them to weigh the importance of different words in an input sequence and determine their influence on the output, regardless of their position. This ability is crucial for understanding context and relationships within language.

C is the correct answer

Why not the others? A. Transformer-based language models use convolutional layers... This describes Convolutional Neural Networks (CNNs), not transformers. B. Transformer-based language models can process only text data. Transformers can be applied to text, images, audio, video, and multimodal data. D. Transformer-based language models process data sequences one element at a time in cyclic iterations. This describes Recurrent Neural Networks (RNNs), which process sequences sequentially. Transformers process tokens in parallel using attention.

Self-attention is the core innovation behind transformer architectures