Better and Worse with Scale: How Contextual Entrainment Diverges with Model Size
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866917409447739392 |
|---|---|
| author | Kukreja, Dikshant Sah, Kshitij Gupta, Gautam Anand, Avinash Shah, Rajiv Ratn Wang, Zhengkui Ng, Aik Beng Cambria, Erik |
| author_facet | Kukreja, Dikshant Sah, Kshitij Gupta, Gautam Anand, Avinash Shah, Rajiv Ratn Wang, Zhengkui Ng, Aik Beng Cambria, Erik |
| contents | Larger language models become simultaneously better and worse at handling contextual information -- better at ignoring false claims, worse at ignoring irrelevant tokens. We formalize this apparent paradox through the first scaling laws for contextual entrainment, the tendency of models to favor tokens that appeared in context regardless of relevance. Analyzing the Cerebras-GPT (111M-13B) and Pythia (410M-12B) model families, we find entrainment follows predictable power-law scaling, but with opposite trends depending on context type: semantic contexts show decreasing entrainment with scale, while non-semantic contexts show increasing entrainment. Concretely, the largest models are four times more resistant to counterfactual misinformation than the smallest, yet simultaneously twice as prone to copying arbitrary tokens. These diverging trends, which replicate across model families, suggest that semantic filtering and mechanical copying are functionally distinct behaviors that scale in opposition -- scaling alone does not resolve context sensitivity, it reshapes it. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_13275 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Better and Worse with Scale: How Contextual Entrainment Diverges with Model Size Kukreja, Dikshant Sah, Kshitij Gupta, Gautam Anand, Avinash Shah, Rajiv Ratn Wang, Zhengkui Ng, Aik Beng Cambria, Erik Computation and Language Machine Learning I.2.7 Larger language models become simultaneously better and worse at handling contextual information -- better at ignoring false claims, worse at ignoring irrelevant tokens. We formalize this apparent paradox through the first scaling laws for contextual entrainment, the tendency of models to favor tokens that appeared in context regardless of relevance. Analyzing the Cerebras-GPT (111M-13B) and Pythia (410M-12B) model families, we find entrainment follows predictable power-law scaling, but with opposite trends depending on context type: semantic contexts show decreasing entrainment with scale, while non-semantic contexts show increasing entrainment. Concretely, the largest models are four times more resistant to counterfactual misinformation than the smallest, yet simultaneously twice as prone to copying arbitrary tokens. These diverging trends, which replicate across model families, suggest that semantic filtering and mechanical copying are functionally distinct behaviors that scale in opposition -- scaling alone does not resolve context sensitivity, it reshapes it. |
| title | Better and Worse with Scale: How Contextual Entrainment Diverges with Model Size |
| topic | Computation and Language Machine Learning I.2.7 |
| url | https://arxiv.org/abs/2604.13275 |