基本信息
浏览量:88
职业迁徙
个人简介
I used to study representations in democracies (almost did a PhD in political science), and now I study representations of language and cognition. My current research focuses on two directions:
Instruction-tuned models. I co/first-authored various large language models (T0, Flan-T5/PaLM, and BLOOM), with a focus on zero-shot generalization to unseen tasks that go beyond statistical pattern matching. At the moment, I’m working on LM agents with recurrent and adaptive computation with reinforcement learning.
Finding where in pretraining and instruction-tuning corpora do models acquire zero-shot and few-shot abilities, understanding exactly how models generalize at test time, and explaining why models match human behaviors in profound ways in some settings (e.g., Dasgupta et al. 2022), while also exhibiting comically un-human-like behaviors in some other settings (e.g., Webson et al. 2023).
Instruction-tuned models. I co/first-authored various large language models (T0, Flan-T5/PaLM, and BLOOM), with a focus on zero-shot generalization to unseen tasks that go beyond statistical pattern matching. At the moment, I’m working on LM agents with recurrent and adaptive computation with reinforcement learning.
Finding where in pretraining and instruction-tuning corpora do models acquire zero-shot and few-shot abilities, understanding exactly how models generalize at test time, and explaining why models match human behaviors in profound ways in some settings (e.g., Dasgupta et al. 2022), while also exhibiting comically un-human-like behaviors in some other settings (e.g., Webson et al. 2023).
研究兴趣
论文共 18 篇作者统计合作学者相似作者
按年份排序按引用量排序主题筛选期刊级别筛选合作者筛选合作机构筛选
时间
引用量
主题
期刊级别
合作者
合作机构
Tao Tu, Anil Palepu,Mike Schaekermann, Khaled Saab,Jan Freyberg,Ryutaro Tanno,Amy Wang, Brenna Li,Mohamed Amin,Nenad Tomasev,Shekoofeh Azizi,Karan Singhal,
CoRR (2024)
引用0浏览0EI引用
0
0
arxiv(2023)
引用6浏览0EI引用
6
0
Sheng Shen, Lei Hou,Yanqi Zhou, Nan Du,Shayne Longpre, Jason Lee,Hyung Won Chung,Barret Zoph,William Fedus,Xinyun Chen,Tu Vu,Yuexin Wu,
arXiv (Cornell University) (2023)
引用0浏览0引用
0
0
加载更多
作者统计
合作学者
合作机构
D-Core
- 合作者
- 学生
- 导师
数据免责声明
页面数据均来自互联网公开来源、合作出版商和通过AI技术自动分析结果,我们不对页面数据的有效性、准确性、正确性、可靠性、完整性和及时性做出任何承诺和保证。若有疑问,可以通过电子邮件方式联系我们:report@aminer.cn