Prompt engineering and model choice yield only marginal gains in LLM rating prediction from text, with the primary ceiling determined by the information content of the input itself.
GPT as a Measurement Tool
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
-
The signal is the ceiling: Measurement limits of LLM-predicted experience ratings from open-ended survey text
Prompt engineering and model choice yield only marginal gains in LLM rating prediction from text, with the primary ceiling determined by the information content of the input itself.