← Search

Xiaotang Du

2 accepted papers

2025

Are We Done with MMLU?

NAACL 2025long

Maybe not. We identify and analyse errors in the popular Massive Multitask Language Understanding (MMLU) benchmark. Even though MMLU is widely adopted, our analysis demonstrates numerous ground truth errors that obscure the true capabilities of LLMs. For example, we find that 57% of the analysed que…

2025

Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering

NAACL 2025long

Large language models (LLMs) can store a significant amount of factual knowledge in their parameters. However, their parametric knowledge may conflict with the information provided in the context—this phenomenon, known as context-memory knowledge conflicts, can lead to undesirable model behaviour, s…