List your company

Confidence Calibration on LogiQA (out-of-distribution)

8ECEADVICE w/ ConfTuner
6.16818.53430.943.266Oct 13, 2025

Evaluation results

MethodLinks
ADVICE w/ ConfTunerModel=GEMMA2 9B-IT2025.1086.224.669
ADVICEModel=GEMMA2 9B-IT2025.1011.910.925.857.8
ConfTunerModel=GEMMA2 9B-IT2025.1018.417.923.871.4
ADVICEModel=LLAMA3.1 8B INSTRUCT2025.102321.230.157.9
ADVICE w/ ConfTunerModel=MISTRAL 7B INSTRUCT2025.102423.128.168
ConfTunerModel=MISTRAL 7B INSTRUCT2025.1024.824.822.163.1
ADVICE w/ ConfTunerModel=LLAMA3.1 8B INSTRUCT2025.102623.831.158.6
ConfTunerModel=LLAMA3.1 8B INSTRUCT2025.1028.628.232.554.4
ADVICEModel=MISTRAL 7B INSTRUCT2025.1035.631.638.259.3
PromptingModel=GEMMA2 9B-IT2025.103938.740.350.3
Self-ConsistencyModel=GEMMA2 9B-IT2025.103938.841.844.8
DefaultModel=GEMMA2 9B-IT2025.1039.13940.450.9
Self-ConsistencyModel=LLAMA3.1 8B INSTRUCT2025.1045.345.146.245.6
Self-ConsistencyModel=MISTRAL 7B INSTRUCT2025.1045.745.443.259
PromptingModel=LLAMA3.1 8B INSTRUCT2025.1048.147.44850.4
PromptingModel=MISTRAL 7B INSTRUCT2025.1049.24849.251.4
DefaultModel=MISTRAL 7B INSTRUCT2025.1051.850.951.352.1
DefaultModel=LLAMA3.1 8B INSTRUCT2025.1053.853.352.950.5