List your company

Arithmetic Reasoning on ARITH 2-Choice

12.4Δ AccuracyRBCorr
2.8325.3167.810.284Feb 12, 2026

Evaluation results

MethodLinks
RBCorrModel=Llama3.1-8B, Few-shot prompt format=Yes2026.0212.4——-0.194
Batch Calibration (BC)Model=Llama3.1-8B, Few-shot prompt format=Yes2026.0212.1——-0.086
Contextual Calibration (CC)Model=Llama3.1-8B, Few-shot prompt format=Yes2026.0211.8——-0.182
RBCorrModel=Llama3.1-8B-Instruct, Few-shot prompt format=Yes2026.025.8——-0.151
Batch Calibration (BC)Model=Llama3.1-8B-Instruct, Few-shot prompt format=Yes2026.025.2——-0.09
Contextual Calibration (CC)Model=Llama3.1-8B-Instruct, Few-shot prompt format=Yes2026.023.2——-0.058
BaselineModel=Llama3.1-8B, Few-shot prompt format=Yes2026.02—74.20.203—
BaselineModel=Llama3.1-8B-Instruct, Few-shot prompt format=Yes2026.02—78.80.159—