List your company

Script Confusion Mitigation on FLEURS zh-sim (test)

93Normalized Edit Similarity Accuracyno-prompt
58.6867.5976.585.41Jan 6, 2026

Evaluation results

MethodLinks
no-promptModel Size=large2026.0193
no-promptModel Size=v32026.0193
promptModel Size=large2026.0193
promptModel Size=v22026.0193
promptModel Size=v32026.0193
steerModel Size=large2026.0193
steerModel Size=v22026.0193
steerModel Size=v32026.0193
promptModel Size=medium2026.0191
steerModel Size=medium2026.0191
no-promptModel Size=medium2026.0187
promptModel Size=small2026.0187
steerModel Size=small2026.0187
no-promptModel Size=v22026.0185
no-promptModel Size=small2026.0179
promptModel Size=base2026.0178
steerModel Size=base2026.0178
promptModel Size=tiny2026.0170
steerModel Size=tiny2026.0169
no-promptModel Size=base2026.0167
no-promptModel Size=tiny2026.0160