List your company

Meta-evaluation of Trajectory Evaluation Metrics on SGRL Benchmarks Aggregate

95.8Oracle Rank Accuracy (Acc)Interval-Paired Preference (IPP)
9.4831.8954.376.71Jun 16, 2026

Evaluation results

MethodLinks
Interval-Paired Preference (IPP)2026.0695.837.762.578812.8335.0953.3473.7600
Lexicographic Return (LR)2026.0695.637.561.68385033.966.1677.8100
Return-Paired Preference (RPP)2026.0694.237.563.283851.8834.8261.5178.3500
Success weighted by Path Length (SPL)2026.0691.831.648.171705.563.4239.0260.200
Success Rate (SR)2026.0612.8007573074.944.9858.4700
Path Reward (PR)2026.0612.80082822.1449.7156.5373.3700