Rationale for Cumulative Objective in Dialogue Models
A simplified, but flawed, approach to training a multi-round dialogue model might be to only maximize the probability of the final response in a conversation, conditioned on the entire history. Explain why the standard training objective, which sums the log-probabilities of every model response in the conversation, is a more effective approach for developing a capable dialogue agent.
0
1
Tags
Ch.4 Alignment - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Analysis in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Comparison of Training Implementations for Multi-Round Dialogue Models
A team is developing a chatbot for multi-turn conversations. They have a dataset of K-round dialogues, each consisting of a sequence of user inputs (x^k) and desired model responses (y^k). To train the model, they must define an objective function that the model's parameters will be optimized to maximize. Which of the following objective functions correctly represents the goal of making the model generate the entire sequence of desired responses accurately within the conversational context?
Diagnosing Context-Loss in a Dialogue Model
Rationale for Cumulative Objective in Dialogue Models
Dataset-Level Objective for Multi-Round Conversational Models