How can we understand neural chatbots in terms of interpretable, symbolic mechanisms? To explore this question, we constructed a Transformer that implements the classic ELIZA chatbot algorithm (with @Abhishek_034 and @danqi_chen). Paper: arxiv.org/abs/2407.10949 (1/6)
Latest posts 5
I'll be presenting this project at 10:30 this morning in La Nouvelle Orleans Ballroom and at the poster session from 10:45 to 12:45 (#1509) - come by! #NeurIPS2023
Learning Transformer ProgramsView quoted post on X
We designed a modified Transformer that can be trained to solve a task and then automatically converted into a discrete, human-readable program. With @_awettig and @danqi_chen.
Paper: arxiv.org/abs/2306.01128
Code: github.com/princeton-nlp/…
[1/12]
We often interpret neural nets by studying simplified representations (e.g. low-dim visualization). But how faithful are these simplifications to the original model? In our new preprint, we found some surprising "interpretability illusions"... 1/6
Learning Transformer Programs
We designed a modified Transformer that can be trained to solve a task and then automatically converted into a discrete, human-readable program. With @_awettig and @danqi_chen.
Paper: arxiv.org/abs/2306.01128
Code: github.com/princeton-nlp/…
[1/12]
We designed a modified Transformer that can be trained to solve a task and then automatically converted into a discrete, human-readable program. With @_awettig and @danqi_chen.
Paper: arxiv.org/abs/2306.01128
Code: github.com/princeton-nlp/…
[1/12]
Many NLP datasets contain unwanted shortcuts. How can we discover these automatically? We propose an approach in “Finding Datasets with Grammar Induction”, to appear at EMNLP ’22 (with @_awettig & @danqi_chen). [1/6]
Paper: arxiv.org/abs/2210.11560
Code: github.com/princeton-nlp/…
Paper: arxiv.org/abs/2210.11560
Code: github.com/princeton-nlp/…