Back to glossary

Transformer architecture

Transformer mimarisi

D2

The Transformer is a neural-network architecture that relies on attention mechanisms to model relationships between elements of input and output sequences.

Review status: 2027-08-28

Technical explanation

In the original sequence-transduction design, self-attention relates positions within a sequence and the architecture avoids sequence-aligned recurrence and convolution.

Conceptual boundaries

A Transformer is an architecture, not a synonym for a large language model, and a language model can be implemented with a particular architecture.

Provider-neutral example

A system can use a Transformer-based model to relate words across a support request before producing a proposed summary.

Limitations

The original paper reports results for stated tasks and configurations, so its architecture description is not a universal performance promise.

Related concepts

Atomic claims and evidence

  1. 1.1Vaswani et al., Attention Is All You Need
    Source
    Vaswani et al., Attention Is All You Need
    Source role
    Authoritative source
    Exact locator
    Abstract; Section 1, paragraphs 1 and 4
    Supported claim
    The original Transformer architecture is based solely on attention mechanisms and dispenses with recurrence and convolutions.
    Last verification
    Review due
    Scope limitation
    This describes the architecture proposed in the cited paper, not every later model called a Transformer.
  2. 2.1Vaswani et al., Attention Is All You Need
    Source
    Vaswani et al., Attention Is All You Need
    Source role
    Authoritative source
    Exact locator
    Section 2, Self-Attention paragraph
    Supported claim
    Self-attention relates different positions of one sequence to compute a representation of that sequence.
    Last verification
    Review due
    Scope limitation
    This is the paper’s account of self-attention and does not imply that every attention mechanism has the same form.