List your company

Script Confusion Mitigation on FLEURS zh-trad (test)

91Accuracyprompt
61.8869.447784.56Jan 6, 2026

Evaluation results

MethodLinks
promptModel Size=v22026.0191
steerModel Size=v22026.0191
promptModel Size=medium2026.0190
steerModel Size=medium2026.0189
steerModel Size=large2026.0189
steerModel Size=v32026.0189
promptModel Size=small2026.0186
steerModel Size=small2026.0186
promptModel Size=base2026.0179
promptModel Size=large2026.0179
promptModel Size=v32026.0179
steerModel Size=base2026.0179
no-promptModel Size=v22026.0173
promptModel Size=tiny2026.0171
steerModel Size=tiny2026.0171
no-promptModel Size=base2026.0169
no-promptModel Size=small2026.0169
no-promptModel Size=medium2026.0169
no-promptModel Size=large2026.0165
no-promptModel Size=v32026.0165
no-promptModel Size=tiny2026.0163