Timezone: »

Describing Differences between Text Distributions with Natural Language
Ruiqi Zhong · Charlie Snell · Dan Klein · Jacob Steinhardt

Tue Jul 19 02:05 PM -- 02:10 PM (PDT) @ Room 301 - 303
How do two \textit{distributions} of text differ?Humans are slow at answering this, since discovering patterns might require tediously reading through hundreds of samples.We propose to automatically summarize the differences by ``learning a natural language hypothesis":given two distributions $D_{0}$ and $D_{1}$, we search for a description that is more often true for $D_{1}$, e.g., ``\textit{is military-related.}"To tackle this problem, we fine-tune GPT-3 to propose descriptions with the prompt: ``[samples of $D_{0}$] + [samples of $D_{1}$] + \textit{the difference between them is \underline{\space\space\space\space}}".We then re-rank the descriptions by checking how often they hold on a larger set of samples with a learned verifier.On a benchmark of 54 real-world binary classification tasks, while GPT-3 Curie (13B) only generates a description similar to human annotation 7\% of the time, the performance reaches 61\% with fine-tuning and re-ranking, and our best system using GPT-3 Davinci (175B) reaches 76\%.We apply our system to describe distribution shifts, debug dataset shortcuts, summarize unknown tasks, and label text clusters, and present analyses based on automatically generated descriptions.

Author Information

Ruiqi Zhong (UC Berkeley)
Charlie Snell (UC Berkeley)
Dan Klein (UC Berkeley)
Jacob Steinhardt (UC Berkeley)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors