A personalized calibration network that combines LLM answers to multiple rubric questions predicted human judges' overall satisfaction scores on dialogues about twice as accurately as the uncalibrated LLM.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
LLM-Rubric: A Multidimensional, Calibrated Approach to Automated Evaluation of Natural Language Texts
A personalized calibration network that combines LLM answers to multiple rubric questions predicted human judges' overall satisfaction scores on dialogues about twice as accurately as the uncalibrated LLM.