List your company

Confidence Calibration on MMLU out-of-distribution

5.6ECEADVICE
4.37212.66120.9529.239Oct 13, 2025

Evaluation results

MethodLinks
ADVICEModel=GEMMA2 9B-IT2025.105.63.918.665.9
ADVICEModel=LLAMA3.1 8B INSTRUCT2025.108.67.620.769.2
ADVICE w/ ConfTunerModel=LLAMA3.1 8B INSTRUCT2025.109.67.120.968.7
ConfTunerModel=GEMMA2 9B-IT2025.10117.320.275.7
ADVICE w/ ConfTunerModel=GEMMA2 9B-IT2025.1011.711.719.966.8
ConfTunerModel=LLAMA3.1 8B INSTRUCT2025.1013.913.924.258.2
DefaultModel=GEMMA2 9B-IT2025.10212124.750.1
PromptingModel=GEMMA2 9B-IT2025.10212124.650.4
ADVICE w/ ConfTunerModel=MISTRAL 7B INSTRUCT2025.1021.821.326.569.9
Self-ConsistencyModel=GEMMA2 9B-IT2025.1022.521.825.857.4
Self-ConsistencyModel=LLAMA3.1 8B INSTRUCT2025.10252529.953.9
PromptingModel=LLAMA3.1 8B INSTRUCT2025.1025.525.328.552.8
DefaultModel=LLAMA3.1 8B INSTRUCT2025.1026.926.729.750.8
ADVICEModel=MISTRAL 7B INSTRUCT2025.1028.727.331.462.7
Self-ConsistencyModel=MISTRAL 7B INSTRUCT2025.1034.534.336.161
DefaultModel=MISTRAL 7B INSTRUCT2025.1035.3353751.5
ConfTunerModel=MISTRAL 7B INSTRUCT2025.10363634.370.8
PromptingModel=MISTRAL 7B INSTRUCT2025.1036.334.138.249.9