List your company

Judgment Consistency on CoinFlip (unseen)

51.8Baseline ScoreUNWAVERING-FQ (+ SFT + DPO)
48.47249.33650.251.064Oct 3, 2023

Evaluation results

MethodLinks
UNWAVERING-FQ (+ SFT + DPO)Type=O2023.1051.818.235.14
Vicuna-7B + SFTType=L2023.1051.41835.02
UNWAVERING-FQ (+ SFT + DPO)Type=L2023.1050.827.253.54
Vicuna-7B + SFTType=C2023.1050.62.85.53
Vicuna-7B + SFTType=O2023.1050.637.273.52
UNWAVERING-FQ (+ SFT + DPO)Type=C2023.1050.40.20.4
Vicuna-7BType=C2023.1050.200
Vicuna-7BType=O2023.104949100
Vicuna-7BType=L2023.1048.61734.98