List your company

Reward Modeling on HHH-Alignment Reversed

86.2AccuracyICRM
5.0826.1447.268.26Feb 9, 2026

Evaluation results

MethodLinks
ICRMN (Number of in-context demonstrations)=322026.0286.2
ICRMN (Number of in-context demonstrations)=162026.0285.4
ICRMN (Number of in-context demonstrations)=82026.0284.2
ICRMN (Number of in-context demonstrations)=42026.0275.1
ICRMN (Number of in-context demonstrations)=22026.0273.2
ICRMN (Number of in-context demonstrations)=12026.0262.7
BTRM2026.0213.6
GRM2026.0211.8
ArmoRM2026.028.2