For a long time, I have had the habit of scanning Arxiv daily to keep up with the latest results in the field and to remind myself that "not to advance is to go back." Many readers have previously asked me how I browse Arxiv and what helper tools I use. In fact, for a long time, I directly browsed the Arxiv official website without using any algorithmic filtering—I went through them one by one myself. This process is tedious, but not unacceptable. The reason for not using algorithms for initial screening was mainly the concern about missed recalls; after all, "scanning" is for chasing the new, and once an algorithm misses a recall, one "loses the first-mover advantage."
Since the release of Kimi Chat, I had been planning to write a helper website combined with Kimi to accelerate the process of scanning papers. Having been a bit more free in the last few weeks, with the help of GPT-4 and Kimi, I have preliminarily completed this website. After several days of testing and optimization, it has gradually become stable, so I am formally inviting readers to try it out.
Writing in the Middle #
As the name "Cool Papers" suggests, this website hopes to make browsing papers a very "cool" and immersive experience. Of course, the current implementation is still quite rudimentary. The "coolness" is mainly reflected in the fact that it answers several FAQ questions for a paper through Kimi. This allows us to more accurately and efficiently understand the main content of a paper (compared to only looking at the title and abstract), thereby judging whether it is a paper that needs deep reading.
Particular points to note:
This is a website for "browsing" papers, not "reading" papers. "Browsing" means "screening" and "filtering." The purpose of "browsing" is to find papers that need intensive reading, rather than replacing intensive reading;
Currently, it only supports Arxiv as a paper source, synchronizing the latest day's paper list from Arxiv. Therefore, browsing Cool Papers is essentially equivalent to browsing Arxiv. Other paper sources, such as OpenReview, might be connected in the future, depending on subsequent usage and feedback;
Because it is positioned for "browsing," "persistence is key, and it waits for no one." Therefore, it currently only supports displaying the latest day's papers and does not temporarily support historical backtracking. This can also be adjusted based on readers' subsequent feedback and needs;
The FAQs are based on Kimi Chat; please be grateful and cherish it. In fact, based on the title and abstract, many papers can already be screened out. The existence of Kimi FAQ is for making more accurate judgments on uncertain papers, so please do not click "[Kimi]" indiscriminately;
Clicking "[PDF]" allows for a preview of the paper's content (limited to PC browsers; mobile browsers will trigger a download). However, this depends on your own network connection to access Arxiv, so if the PDF preview takes a long time to appear, it may be a local network issue;
Clicking "[Copy]" will copy the basic information of the paper (title, abstract, link, etc.) to the clipboard, allowing it to be pasted elsewhere to share the paper;
The paper list maintains the publication order of Arxiv by default. If "-sorted-by-stars" is added, it will be sorted according to stars calculated based on the clicks of all users;
Paper Updates: Paper updates are directly synchronized from the Arxiv official website, normally with a delay of no more than 10 minutes. Arxiv updates generally occur around 10:00 AM (Beijing time) on weekdays, but there may be fluctuations of several hours, and updates do not occur on Saturdays, Sundays, or some US holidays. If you find that the website papers have not updated, you can check the update status on the official Arxiv site.
Further Explanation regarding Kimi FAQ:
After clicking "[Kimi]" for each paper, you will enter a queue;
If there are many people in the queue, it will display a "Pending:xxx" result, where "xxx" is the queue count; this number will update automatically;
Once the queuing is complete, it will display "Loading:xxx%", followed by a streaming output of the FAQ content;
The queuing and generation process does not require maintaining the page open; even if you close the page, the queuing and generation will continue in the background. Reopening the page and clicking the same "[Kimi]" will restore the original progress;
If you find that the FAQ content is output instantly after clicking "[Kimi]," it means the paper has already been read by another reader and the FAQ has been cached;
Because there is a queuing mechanism, clicking "[Kimi]" indiscriminately won't put too much pressure on the website, but it will make others' waiting times too long, which is unfriendly behavior.
Writing at the End #
Finally, everyone's opinions and suggestions are welcome! Use cases and demands are also welcome. Obviously, the current Cool Papers is still very rough and far from being as "cool" as imagined. Using Kimi for FAQs is only a very basic scheme for combining Kimi with papers. Kimi's ultra-long context window should offer much greater imaginative space. Therefore, I look forward to everyone giving play to their imagination to find a more perfect way to combine Kimi with paper browsing.