Papers by Kshitij Shah
Better and Worse with Scale: How Contextual Entrainment Diverges with Model Size (2026.findings-acl)
Copied to clipboard
Dikshant Kukreja, Kshitij Sah, Gautam Gupta, Avinash Anand, Rajiv Ratn Shah, Zhengkui Wang, Aik Beng Ng, Erik Cambria
| Challenge: | Larger language models become better and worse at handling contextual information . et al. (2017) formalized contextual entrainment as a tendency to favor tokens in context . |
| Approach: | They formalize the first scaling laws for contextual entrainment . they find large models are four times more resistant to counterfactual misinformation . |
| Outcome: | The largest models are four times more resistant to counterfactual misinformation than the smallest, but twice as prone to copying arbitrary tokens. |
Correcting the Autocorrect: Context-Aware Typographical Error Correction via Training Data Augmentation (2020.lrec-1)
Copied to clipboard
| Challenge: | a recent study shows that typographical errors are now ubiquitous . traditional spelling correction software is inadequate to correct typographical mistakes . |
| Approach: | They propose to generate typographical errors based on annotated spelling errors . they then use annotations to introduce errors into substantially larger corpora . |
| Outcome: | The proposed method generates typographical errors that require context-aware error detection . it also shows that machine learning can correct typographical mistakes based on the data . |