Learn Before
Multiple Choice

A company is training a language model to act as an automated assistant for processing loan applications. The model must follow a specific, legally-mandated, multi-step procedure to ensure fairness and compliance (e.g., checking credit history, verifying income, providing specific disclosures). The company decides to train the model using a system that provides a large positive reward only if the final loan decision (approve/deny) is correct based on the applicant's overall profile. What is the most significant weakness of this training strategy?

0

1

Updated 2025-10-07

Contributors are:

Who are from:

Tags

Ch.5 Inference - Foundations of Large Language Models

Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences

Evaluation in Bloom's Taxonomy

Cognitive Psychology

Psychology

Social Science

Empirical Science

Science