Is it possible to extract the internal physics knowledge of a large language model as semantic network?ΒΆ
The answer of this question depends on the answer of the following questions:
How can general knowledge be represented as a semantic network?
What questions to ask a LLM to generate a semantic network?
How can one evaluate the result?
What are the peculiarities of physics knowledge?
The following approaches can be used for extracting the physics knowledge from a LLM:
Building a subtopic network