List your company

Confidence Calibration on SciQ (test)

1.5ECESelf-Consistency
0.9684.5598.1511.741Oct 13, 2025

Evaluation results

MethodLinks
Self-ConsistencyModel=LLAMA3.1 8B INSTRUCT2025.101.51.37.174
DefaultModel=LLAMA3.1 8B INSTRUCT2025.103.72.18.252
Self-ConsistencyModel=GEMMA2 9B-IT2025.10446.370.3
PromptingModel=LLAMA3.1 8B INSTRUCT2025.104.30.77.251.4
DefaultModel=GEMMA2 9B-IT2025.104.91.55.250.3
PromptingModel=GEMMA2 9B-IT2025.105.11.45.250
ConfTunerModel=GEMMA2 9B-IT2025.106.82.96.977.2
ADVICEModel=LLAMA3.1 8B INSTRUCT2025.107.97.76.673.4
ADVICEModel=MISTRAL 7B INSTRUCT2025.108.38.112.563
ConfTunerModel=LLAMA3.1 8B INSTRUCT2025.109.59.19.558.4
PromptingModel=MISTRAL 7B INSTRUCT2025.109.88.514.951.7
DefaultModel=MISTRAL 7B INSTRUCT2025.1010.3915.350.3
ConfTunerModel=MISTRAL 7B INSTRUCT2025.1010.89.91468.6
ADVICEModel=GEMMA2 9B-IT2025.1014.214.27.472.6
Self-ConsistencyModel=MISTRAL 7B INSTRUCT2025.1014.814.516.967.1