List your company

Script Confusion Mitigation on FLEURS sr-latn (test)

96Accuracy (Normalized Edit Similarity)steer
61.6870.5979.588.41Jan 6, 2026

Evaluation results

MethodLinks
steerModel Size=large2026.0196
steerModel Size=v22026.0196
steerModel Size=v32026.0196
promptModel Size=large2026.0195
promptModel Size=v22026.0195
promptModel Size=v32026.0195
no-promptModel Size=large2026.0193
no-promptModel Size=v32026.0193
promptModel Size=medium2026.0193
steerModel Size=medium2026.0193
no-promptModel Size=small2026.0190
promptModel Size=small2026.0190
steerModel Size=small2026.0189
no-promptModel Size=base2026.0181
no-promptModel Size=v22026.0181
promptModel Size=base2026.0181
steerModel Size=base2026.0180
no-promptModel Size=tiny2026.0166
promptModel Size=tiny2026.0164
steerModel Size=tiny2026.0164
no-promptModel Size=medium2026.0163