11 terms
Showing all terms starting with P
The internal numerical weights of an AI model that are learned during training and determine how inputs are transformed into outputs.
The input text or instruction provided to an AI model to guide its response or action.
The practice of crafting and refining prompts to elicit more accurate, relevant, or creative outputs from AI models.
The initial phase of training a large model on a broad, general dataset before fine-tuning it for specific tasks.
A metric measuring how well a language model predicts a text sample - lower perplexity indicates a better, more confident model.
A sequence of automated data processing and modelling steps from raw input to final prediction, enabling reproducible and scalable ML workflows.
In reinforcement learning, a policy is the strategy an agent follows to choose actions based on the current state of the environment.
A downsampling operation in CNNs or embedding models that aggregates local features, reducing spatial dimensions and computation.
Vectors added to token embeddings in transformers to provide sequence order information, since attention itself is position-agnostic.
Precision measures the proportion of positive predictions that are correct; recall measures how many actual positives were correctly identified.
Removing redundant or low-importance weights from a trained model to reduce its size and inference cost with minimal accuracy loss.
