Is a pre-trained model an untrained model for a new task?
No. A pre-trained model already contains learned patterns. It may still need prompting, adaptation, or additional training to perform a new task reliably.
Explore This Question
Search for additional explanations using these tools:
More FAQs in How does an untrained neural network differ from a trained one?
No. Many weights use random initialization, but biases, normalization parameters, or other components may use fixed starting values chosen for the architecture.
Yes. Different random seeds, training examples, data order, and optimization settings can lead networks with the same architecture to different final parameter values.
Still have questions?
Browse all our FAQs or reach out to our support team