List your company

Expected Calibration Error on PopQA

19.85ECECDKC
17.795231.665145.53559.4049Feb 13, 2026

Evaluation results

MethodLinks
CDKCBackbone=Qwen2.5-7B-Instruct, Optimization Rounds=22026.0219.85
CDKCBackbone=Qwen2.5-7B-Instruct2026.0227.91
GRPOBackbone=Qwen2.5-7B-Instruct2026.0230.03
BARRELBackbone=Qwen2.5-7B-Instruct2026.0256.52
Know WhatBackbone=Qwen2.5-7B-Instruct2026.0261.78
Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.0267.07
CGKEBackbone=Qwen2.5-7B-Instruct2026.0267.23
LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.0267.56
Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.0267.63
CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.0271.22