Formal Representation of an Instruction-Tuned LLM
An instruction-fine-tuned Large Language Model can be formally represented as a conditional probability distribution, denoted as . In this mathematical expression, stands for the instruction provided, represents the user's input, and is the corresponding generated model output or response.
0
1
Tags
Ch.4 Alignment - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Related
Two Levels of Generalization in Instruction-Tuned LLMs
Complexity of Generalization due to Instruction and Input Variation
A development team fine-tunes a large language model to be a helpful assistant for summarizing legal documents. They use a large dataset of legal texts and their corresponding summaries. After deployment, they observe the following:
- The model performs exceptionally well when asked to summarize new, unseen legal documents (e.g., contracts, court rulings).
- However, when users give it slightly different instructions, such as 'Explain this legal clause in simple terms,' 'Extract the key dates from this document,' or 'Translate this legal paragraph into French,' the model's performance is poor and unreliable.
Based on this scenario, which statement best analyzes the model's generalization capabilities?
Evaluating Fine-Tuning Strategies for Generalization
Performance Metric for Instruction-Tuned LLMs
Formal Representation of an Instruction-Tuned LLM
A large language model has been fine-tuned on a variety of instructional tasks. Match each of the following performance observations with the specific type of generalization challenge it represents.
Learn After
A user interacts with a language model by providing an instruction and an input. The instruction is 'Summarize the key point of the following text.' The input text is 'Jupiter is the fifth planet from the Sun and the largest in the Solar System.' The model generates the output 'Jupiter is the largest planet in our solar system.' How is the conditional probability of the model generating this specific output formally represented?
An instruction-tuned model's behavior can be formally represented as a conditional probability distribution
Pr(y|c, z). Match each variable from this representation to its corresponding description.Analyzing Model Output Probability