Unit 6 Lecture

Fine-tuning

The process of adjusting pretrained language models to improve performance on specific downstream tasks.

Instruction Fine-Tuning

A method of teaching models to follow natural language instructions, focusing on aligning their behavior with user expectations.

Parameter Efficient Fine-Tuning (PEFT)

A technique that adapts large models to new tasks by training fewer parameters, improving resource efficiency.

Single-task Fine-Tuning

Adapts a pretrained model to excel at one specific task, requiring task-specific labeled datasets.

Multi-task Fine-Tuning

Simultaneously fine-tunes an LLM for multiple tasks, capturing shared knowledge and improving efficiency.

ROUGE

A set of metrics used for evaluating the quality of summaries by comparing them to reference summaries.

BLEU

A metric for evaluating machine translation quality by comparing the overlap between machine-generated and reference translations.

Adapters

Small, trainable modules that can be inserted into a model to facilitate efficient fine-tuning.

LoRA (Low-Rank Adaptation)

A PEFT technique that adds low-rank matrices to model weights, significantly reducing the number of trainable parameters.

Chatbot Personalization

The process of tailoring a chatbot's responses and behavior based on user interactions and preferences.