Explorations of Self-Repair in Language Models
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_ | 1866909579258888192 |
|---|---|
| author | Rushing, Cody Nanda, Neel |
| author_facet | Rushing, Cody Nanda, Neel |
| contents | Prior interpretability research studying narrow distributions has preliminarily identified self-repair, a phenomena where if components in large language models are ablated, later components will change their behavior to compensate. Our work builds off this past literature, demonstrating that self-repair exists on a variety of models families and sizes when ablating individual attention heads on the full training distribution. We further show that on the full training distribution self-repair is imperfect, as the original direct effect of the head is not fully restored, and noisy, since the degree of self-repair varies significantly across different prompts (sometimes overcorrecting beyond the original effect). We highlight two different mechanisms that contribute to self-repair, including changes in the final LayerNorm scaling factor and sparse sets of neurons implementing Anti-Erasure. We additionally discuss the implications of these results for interpretability practitioners and close with a more speculative discussion on the mystery of why self-repair occurs in these models at all, highlighting evidence for the Iterative Inference hypothesis in language models, a framework that predicts self-repair. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_15390 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Explorations of Self-Repair in Language Models Rushing, Cody Nanda, Neel Machine Learning Artificial Intelligence Computation and Language Prior interpretability research studying narrow distributions has preliminarily identified self-repair, a phenomena where if components in large language models are ablated, later components will change their behavior to compensate. Our work builds off this past literature, demonstrating that self-repair exists on a variety of models families and sizes when ablating individual attention heads on the full training distribution. We further show that on the full training distribution self-repair is imperfect, as the original direct effect of the head is not fully restored, and noisy, since the degree of self-repair varies significantly across different prompts (sometimes overcorrecting beyond the original effect). We highlight two different mechanisms that contribute to self-repair, including changes in the final LayerNorm scaling factor and sparse sets of neurons implementing Anti-Erasure. We additionally discuss the implications of these results for interpretability practitioners and close with a more speculative discussion on the mystery of why self-repair occurs in these models at all, highlighting evidence for the Iterative Inference hypothesis in language models, a framework that predicts self-repair. |
| title | Explorations of Self-Repair in Language Models |
| topic | Machine Learning Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2402.15390 |