GPT-4o and GPT-4.5-preview reached moderate to substantial Cohen's kappa with human coders for three of four student thinking themes after per-theme hyperparameter and prompt tuning.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
physics.ed-ph 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Investigation of the Inter-Rater Reliability between Large Language Models and Human Raters in Qualitative Analysis
GPT-4o and GPT-4.5-preview reached moderate to substantial Cohen's kappa with human coders for three of four student thinking themes after per-theme hyperparameter and prompt tuning.