FAQ

Here are some frequently asked questions about Mind The Abstract's features and how to customize your experience:

How are summaries generated?

Summaries are generated by our custom agentic system that leverages predominantly small language models (SLMs). We are constantly working on improving our summaries, so you may notice how summaries are written changes over time. This largely depends on our context engineering and choice of SLMs to power our agents.

Can I change the categories of my digest?

Yes! The end of each email contians a link to manage your preferences. We currently offer the following arXiv categories:

• Artifical Inteligence (cs.AI)
• Computer Vision (cs.CV)
• Computation and Language (cs.CL)
• Machine Learning (cs.LG and stat.ML)
• Neural Evolutionary Computing (cs.NE)

As the newsletter grows, we plan to add more categories for you to choose from!

The summaries are too long and informal, are there any alternative summary types?

In addition to our default, informal summary flavor we currently offer a "TLDR" version, which is located in user preferences. We recommend the "TLDR" version for researchers or those looking to get the nuts and bolts of a paper ASAP. We offer the "Informal" summary flavor by default, as we feel it's the best way to gain an intuitive sense of what a paper is doing.

How are summaries verified for correctness?

To verify the accuracy of summaries we parse the summaries ourselves and utilize LLMs. If you happen to catch a mistake, please reach out to info@mindtheabstract.com and we'll correct the browser version of the summary ASAP.

Why are these the summaries I am seeing for the previous week of papers on arXiv?

The 10 summaries delievered weekly to your inbox are selected based on our custom clustering algorithm. We try and weigh different factors about the content and writing style of a paper. We end up summarizing the highest scoring papers by prefrence or combination of preferences.