English

Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models

Machine Learning 2024-10-31 v2 Artificial Intelligence Computation and Language

Abstract

What latent features are encoded in language model (LM) representations? Recent work on training sparse autoencoders (SAEs) to disentangle interpretable features in LM representations has shown significant promise. However, evaluating the quality of these SAEs is difficult because we lack a ground-truth collection of interpretable features that we expect good SAEs to recover. We thus propose to measure progress in interpretable dictionary learning by working in the setting of LMs trained on chess and Othello transcripts. These settings carry natural collections of interpretable features -- for example, "there is a knight on F3" -- which we leverage into supervised\textit{supervised} metrics for SAE quality. To guide progress in interpretable dictionary learning, we introduce a new SAE training technique, p-annealing\textit{p-annealing}, which improves performance on prior unsupervised metrics as well as our new metrics.

Keywords

Cite

@article{arxiv.2408.00113,
  title  = {Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models},
  author = {Adam Karvonen and Benjamin Wright and Can Rager and Rico Angell and Jannik Brinkmann and Logan Smith and Claudio Mayrink Verdun and David Bau and Samuel Marks},
  journal= {arXiv preprint arXiv:2408.00113},
  year   = {2024}
}

Comments

Accepted as an oral paper (top 5%) at the ICML 2024 Mechanistic Interpretability Workshop and to the NeurIPS 2024 Main Conference

R2 v1 2026-06-28T17:59:48.324Z