Improving Dictionary Learning with Gated Sparse Autoencoders
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916229312151552 |
|---|---|
| author | Rajamanoharan, Senthooran Conmy, Arthur Smith, Lewis Lieberum, Tom Varma, Vikrant Kramár, János Shah, Rohin Nanda, Neel |
| author_facet | Rajamanoharan, Senthooran Conmy, Arthur Smith, Lewis Lieberum, Tom Varma, Vikrant Kramár, János Shah, Rohin Nanda, Neel |
| contents | Recent work has found that sparse autoencoders (SAEs) are an effective technique for unsupervised discovery of interpretable features in language models' (LMs) activations, by finding sparse, linear reconstructions of LM activations. We introduce the Gated Sparse Autoencoder (Gated SAE), which achieves a Pareto improvement over training with prevailing methods. In SAEs, the L1 penalty used to encourage sparsity introduces many undesirable biases, such as shrinkage -- systematic underestimation of feature activations. The key insight of Gated SAEs is to separate the functionality of (a) determining which directions to use and (b) estimating the magnitudes of those directions: this enables us to apply the L1 penalty only to the former, limiting the scope of undesirable side effects. Through training SAEs on LMs of up to 7B parameters we find that, in typical hyper-parameter ranges, Gated SAEs solve shrinkage, are similarly interpretable, and require half as many firing features to achieve comparable reconstruction fidelity. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_16014 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Improving Dictionary Learning with Gated Sparse Autoencoders Rajamanoharan, Senthooran Conmy, Arthur Smith, Lewis Lieberum, Tom Varma, Vikrant Kramár, János Shah, Rohin Nanda, Neel Machine Learning Artificial Intelligence Recent work has found that sparse autoencoders (SAEs) are an effective technique for unsupervised discovery of interpretable features in language models' (LMs) activations, by finding sparse, linear reconstructions of LM activations. We introduce the Gated Sparse Autoencoder (Gated SAE), which achieves a Pareto improvement over training with prevailing methods. In SAEs, the L1 penalty used to encourage sparsity introduces many undesirable biases, such as shrinkage -- systematic underestimation of feature activations. The key insight of Gated SAEs is to separate the functionality of (a) determining which directions to use and (b) estimating the magnitudes of those directions: this enables us to apply the L1 penalty only to the former, limiting the scope of undesirable side effects. Through training SAEs on LMs of up to 7B parameters we find that, in typical hyper-parameter ranges, Gated SAEs solve shrinkage, are similarly interpretable, and require half as many firing features to achieve comparable reconstruction fidelity. |
| title | Improving Dictionary Learning with Gated Sparse Autoencoders |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2404.16014 |