Are Language Models Script-Aware?

LLMs and SLMs demonstrate high script fidelity, achieving over 98% accuracy when adapting output to match input scripts or following explicit script instructions. Testing across multi-scriptic languages shows that while both model classes possess significant script knowledge, larger models outperform smaller ones in instruction following.