When evaluation items admit multiple valid responses, gold-label accuracy underestimates true model performance, and the paper offers bounds on the true performance from partial knowledge.
Holistic evaluation of language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Framework for Evaluating LLMs Under Task Indeterminacy
When evaluation items admit multiple valid responses, gold-label accuracy underestimates true model performance, and the paper offers bounds on the true performance from partial knowledge.