Papers by Kshitij Gupta
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. |
Towards Detecting Political Bias in Hindi News Articles (2022.acl-srw)
Copied to clipboard
| Challenge: | Political propaganda in recent times has been amplified by media news portals through biased reporting, creating untruthful narratives on serious issues . a dataset for this task was not available, therefore we developed a transformer-based transfer learning method to fine-tune the pre-trained network on our data. |
| Approach: | They propose a transformer-based transfer learning method to fine-tune the pre-trained network on the data for this bias detection. |
| Outcome: | The proposed method fine-tunes the pre-trained network on the data to detect political bias in Hindi news articles. |
Aurora-M: Open Source Continual Pre-training for Multilingual Language and Code (2025.coling-industry)
Copied to clipboard
Taishi Nakamura, Mayank Mishra, Simone Tedeschi, Yekun Chai, Jason T. Stillerman, Felix Friedrich, Prateek Yadav, Tanmay Laud, Vu Minh Chien, Terry Yue Zhuo, Diganta Misra, Ben Bogin, Xuan-Son Vu, Marzena Karpinska, Arnav Varma Dantuluri, Wojciech Kusa, Tommaso Furlanello, Rio Yokota, Niklas Muennighoff, Suhas Pai, Tosin Adewumi, Veronika Laippala, Xiaozhe Yao, Adalberto Barbosa Junior, Aleksandr Drozd, Jordan Clive, Kshitij Gupta, Liangyu Chen, Qi Sun, Ken Tsui, Nour Moustafa-Fahmy, Nicolo Monti, Tai Dang, Ziyang Luo, Tien-Tung Bui, Roberto Navigli, Virendra Mehta, Matthew Blumberg, Victor May, Hiep Nguyen, Sampo Pyysalo
| Challenge: | Pretrained language models are integral part of AI applications, but their high computational cost limits accessibility. |
| Approach: | They evaluate Aurora-M, a 15B parameter multilingual open-source model trained on English, Finnish, Hindi, Japanese, Vietnamese, and code. |
| Outcome: | The proposed model outperforms existing models on English, Finnish, Hindi, Japanese, Vietnamese, and code. |