Learn Before
Analyzing a Model's Commonsense Failure
A language model is evaluated on its ability to perform grounded commonsense inference. It is given the following context-event pair and asked to determine if the event is a plausible outcome. Analyze the model's incorrect conclusion and explain the specific commonsense reasoning it failed to apply.
0
1
Tags
Ch.2 Generative Models - 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
A language model is tasked with determining if an event is a plausible outcome of a given context. Consider the following pair:
Context: The chef carefully cracked an egg on the side of a hot, oiled skillet. Event: The egg began to cook and turn white.
The model correctly identifies the event as plausible. Which of the following justifications provides the strongest evidence of sound commonsense reasoning for this conclusion?
Analyzing a Model's Commonsense Failure
A large language model is being trained on a text-pair classification task. The goal is for the model to learn to evaluate if the second sentence (the hypothesis) describes a likely, commonsense outcome of the situation described in the first sentence (the premise). Which of the following text pairs is the most suitable example for this specific training objective?