List your company

Conversational Derailment Prediction on CGA-CMV-Large (test)

70.9AccuracySOTA
69.3469.74570.1570.555May 28, 2026

Evaluation results

MethodLinks
SOTADecision Mechanism=Standard2026.0570.934.369.176.172.33.9
SOTA + selective deferralDecision Mechanism=Selective Deferral2026.0570.926.772.168.470.23.8
SOTA + simulation (average)Decision Mechanism=Simulation (Average)2026.0570.236.268.176.6724
SOTA + simulation (majority)Decision Mechanism=Simulation (Majority)2026.057036.967.776.771.84
SOTA (oracle threshold)Decision Mechanism=Oracle Threshold Selection2026.057026.771.566.8693.7
SOTA + random deferralDecision Mechanism=Random Deferral2026.0569.430.2706969.23.8