no code implementations • 26 Jul 2023 • Chenyan Jia, Michelle S. Lam, Minh Chau Mai, Jeff Hancock, Michael S. Bernstein
Finally, in Study 3, we replicate Study 1 using the democratic attitude model instead of manual labels to test its attitudinal and behavioral impact (N=558), and again find that the feed downranking using the societal objective function reduced partisan animosity (d=. 25).
1 code implementation • 26 May 2023 • Ruibo Liu, Ruixin Yang, Chenyan Jia, Ge Zhang, Denny Zhou, Andrew M. Dai, Diyi Yang, Soroush Vosoughi
Social alignment in AI systems aims to ensure that these models behave according to established societal values.
no code implementations • 1 Jan 2023 • Ruibo Liu, Chenyan Jia, Ge Zhang, Ziyu Zhuang, Tony X Liu, Soroush Vosoughi
We present Second Thought, a new learning paradigm that enables language models (LMs) to re-align with human values.
1 code implementation • ICLR 2022 • Ruibo Liu, Chongyang Gao, Chenyan Jia, Guangxuan Xu, Soroush Vosoughi
The performance of existing text style transfer models is severely limited by the non-parallel datasets on which the models are trained.
no code implementations • Findings (ACL) 2021 • Ruibo Liu, Jason Wei, Chenyan Jia, Soroush Vosoughi
Generating context-aware language that embodies diverse emotions is an important step towards building empathetic NLP systems.
no code implementations • 30 Apr 2021 • Ruibo Liu, Chenyan Jia, Jason Wei, Guangxuan Xu, Lili Wang, Soroush Vosoughi
Current large-scale language models can be politically biased as a result of the data they are trained on, potentially causing serious problems when they are deployed in real-world settings.
no code implementations • 5 Jan 2021 • Ruibo Liu, Lili Wang, Chenyan Jia, Soroush Vosoughi
To detect polar words, we train a multi-attribute-aware word embedding model that is aware of ideology and topics on 360k full-length media articles.
no code implementations • EMNLP 2020 • Ruibo Liu, Guangxuan Xu, Chenyan Jia, Weicheng Ma, Lili Wang, Soroush Vosoughi
For instance, Data Boost improves F1 for the three tasks by 8. 7% on average when given only 10% of the whole data for training.