@inproceedings{gur-arieh-etal-2025-enhancing,
    title = "Enhancing Automated Interpretability with Output-Centric Feature Descriptions",
    author = "Gur-Arieh, Yoav  and
      Mayan, Roy  and
      Agassy, Chen  and
      Geiger, Atticus  and
      Geva, Mor",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingestion-acl-25/2025.acl-long.288/",
    pages = "5757--5778",
    ISBN = "979-8-89176-251-0",
    abstract = "Automated interpretability pipelines generate natural language descriptions for the concepts represented by features in large language models (LLMs), such as ``plants'' or ``the first word in a sentence''. These descriptions are derived using inputs that activate the feature, which may be a dimension or a direction in the model{'}s representation space. However, identifying activating inputs is costly, and the mechanistic role of a feature in model behavior is determined both by how inputs cause a feature to activate and by how feature activation affects outputs. Using steering evaluations, we reveal that current pipelines provide descriptions that fail to capture the causal effect of the feature on outputs. To fix this, we propose efficient, output-centric methods for automatically generating feature descriptions. These methods use the tokens weighted higher after feature stimulation or the highest weight tokens after applying the vocabulary ``unembedding'' head directly to the feature. Our output-centric descriptions better capture the causal effect of a feature on model outputs than input-centric descriptions, but combining the two leads to the best performance on both input and output evaluations. Lastly, we show that output-centric descriptions can be used to find inputs that activate features previously thought to be ``dead''."
}