GPT-4.1 predictions from fan text match self-reported experience ratings within one point 67% of the time but are biased low by one point, interpreted as measuring salient moments versus integrated overall judgment.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
LLM Predictive Scoring and Validation: Inferring Experience Ratings from Unstructured Text
GPT-4.1 predictions from fan text match self-reported experience ratings within one point 67% of the time but are biased low by one point, interpreted as measuring salient moments versus integrated overall judgment.