Order Independence With Finetuning

Katrina Brown, Reid McIlroy

Large language models (LLMs) demonstrate remarkable performance on many NLP tasks, yet often exhibit order dependence: simply reordering semantically identical tokens (e.g., answer…