Not a nuisance but a useful heuristic: Outlier dimensions favor frequent tokens in language models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914072157487104 |
|---|---|
| author | Macocco, Iuri Graichen, Nora Boleda, Gemma Baroni, Marco |
| author_facet | Macocco, Iuri Graichen, Nora Boleda, Gemma Baroni, Marco |
| contents | We study last-layer outlier dimensions, i.e. dimensions that display extreme activations for the majority of inputs. We show that outlier dimensions arise in many different modern language models, and trace their function back to the heuristic of constantly predicting frequent words. We further show how a model can block this heuristic when it is not contextually appropriate, by assigning a counterbalancing weight mass to the remaining dimensions, and we investigate which model parameters boost outlier dimensions and when they arise during training. We conclude that outlier dimensions are a specialized mechanism discovered by many distinct models to implement a useful token prediction heuristic. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_21718 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Not a nuisance but a useful heuristic: Outlier dimensions favor frequent tokens in language models Macocco, Iuri Graichen, Nora Boleda, Gemma Baroni, Marco Computation and Language Artificial Intelligence I.2.7 We study last-layer outlier dimensions, i.e. dimensions that display extreme activations for the majority of inputs. We show that outlier dimensions arise in many different modern language models, and trace their function back to the heuristic of constantly predicting frequent words. We further show how a model can block this heuristic when it is not contextually appropriate, by assigning a counterbalancing weight mass to the remaining dimensions, and we investigate which model parameters boost outlier dimensions and when they arise during training. We conclude that outlier dimensions are a specialized mechanism discovered by many distinct models to implement a useful token prediction heuristic. |
| title | Not a nuisance but a useful heuristic: Outlier dimensions favor frequent tokens in language models |
| topic | Computation and Language Artificial Intelligence I.2.7 |
| url | https://arxiv.org/abs/2503.21718 |