← Search

Yuchang Cheng

2 accepted papers

2025

Investigating Neurons and Heads in Transformer-based LLMs for Typographical Errors

EMNLP 2025

This paper investigates how LLMs encode inputs with typos. We hypothesize that specific neurons and attention heads recognize typos and fix them internally using local and global contexts. We introduce a method to identify typo neurons and typo heads that work actively when inputs contain typos. Our

Cited by 0SourcePDFScholar
2025

SubRegWeigh: Effective and Efficient Annotation Weighing with Subword Regularization

COLING 2025main

NLP datasets may still contain annotation errors, even when they are manually annotated. Researchers have attempted to develop methods to automatically reduce the adverse effect of errors in datasets. However, existing methods are time-consuming because they require many trained models to detect err…