List your company

Multiple Choice Question Answering on MMLU (Cross-Benchmark Evaluation)

77.4AccuracyReal-target CSFT
53.27259.53665.872.064Apr 27, 2026

Evaluation results

MethodLinks
Real-target CSFTCondition=Real-target CSFT2026.0477.461.6470
Shuffled-targetCondition=Shuffled-target2026.0456.152.366
BaselineCondition=Baseline2026.0454.253.5498