AI Assistant FAQ

What is the Papers AI Assistant?

The Papers AI Assistant is designed to enhance research efficiency by providing real-time, in-depth analysis, summarization, and contextual understanding of scholarly articles, accelerating how users interact with complex academic content.

How does it work?

The AI Assistant uses generative AI to analyze your question and attempts to identify an answer within the context of the currently opened PDF. It leverages the general knowledge stored within large language models with the specific context of the PDF you have opened to interpret and answer your questions.

What can I do with it?

Our goal is to accelerate the pace of research without compromising the integrity of scholarly consumption and review. We think there are a number of ways the AI Assistant can do this, including:

  • Content Summarization: helping extract key insights from papers.
  • Deep-dive Queries: ask specific questions about an article’s contents
  • Terminology Explanation: help explain complex academic jargon or specific terminology used in the article
  • Contextualized Information: provide additional background context on a topic, to help grasp the implications of the research
  • Identify Research Gaps: analyze the paper to identify possible areas for future research
  • Comprehension Bridges: every research team has varying degrees of proficiency for academic language or the actual language of a given paper; this can make papers more understandable, and digestible, to a wider audience
  • Multilingual capability: users can analyze the paper in different languages

And that’s just the beginning. It is our hope in releasing this private beta that we will also learn new ways in which generative AI can be a tool for research.

How accurate is the AI Assistant? Can I trust its answers?

We have designed our AI Assistant to focus on answering specific questions contained within the opened research paper, rather than relying on broad knowledge trained within large language models. This can help mitigate the probability of hallucinations, but there is always the possibility the AI may misinterpret a question or fail to extract the correct context. Accuracy may also vary depending on the topic. We encourage you to refer back to the full text to verify answers, and in many cases our AI Assistant will point you to the specific paragraphs from which an answer was obtained.

As part of the private beta we are seeking feedback on the quality of responses, and encourage you to use the built-in feedback buttons to share your experiences.

Can I use the AI Assistant to analyze a paper that I don’t have in my library?

No, the AI Assistant utilizes the full text of PDFs stored in your own Papers library. If you do not have a copy of the PDF, then the AI Assistant would not have access to text for analysis.

Does the tool support articles in languages other than English?

Yes, the AI Assistant is capable of working with texts in other languages as well as answering questions in other languages.

Can I use the AI Assistant to help write or format my own scholarly articles?

The AI Assistant was not designed to write articles, but rather to help answer questions within existing articles.

Are there any limits to using the AI Assistant?

We may limit or throttle usage to the AI Assistant from time to time, depending on demand and resource availability during the private beta.

How do I share feedback?

Please use the built-in app feedback buttons to share feedback with the team or email [email protected].

What does “Regenerate Answer” do?

Regenerate Answer attempts a different approach to answer your question. For example, it may modify the context to focus on your most recent question. We encourage you to use the feedback buttons to indicate if a regenerated answer is better or worse.

Can I use the AI Assistant across all my devices?

Currently, the AI Assistant is only available in the web version of Papers. We are still in the early stages of exploring how this might expand in future releases.