Strengths
- The latest AI/ML paper preprints are updated every day
- Completely free, no registration required
- Available before the paper is published (months ahead of the journal)
- Covering CS, mathematics, physics, statistics and other fields
- Provides API and RSS subscriptions
Best for
- Track the latest research progress in the field of AI/ML
- Get a preprint version of your paper
- Learn about the latest work in a certain research direction
- Learn cutting-edge AI technology
- Researchers publish and share work
Efficiently track the latest papers
arXiv publishes a large number of new papers every day, and it is important to master efficient tracking methods.
Subscribe to daily updates in a specific area
Category subscriptions to arXiv: Main AI/ML categories: - cs.AI - Artificial Intelligence - cs.LG - Machine Learning - cs.CV - Computer Vision - cs.CL - Computational Linguistics (NLP) - stat.ML - Statistical Machine Learning How to subscribe: 1. Visit arxiv.org/list/cs.LG/recent 2. Click "Subscribe" to get daily emails 3. Or subscribe to the RSS feed: https://arxiv.org/rss/cs.LG Recommended tools: - Arxiv Sanity Preserver (arxiv-sanity.com) - Papers With Code (paperswithcode.com) - Semantic Scholar's weekly digest
Receive a daily list of the latest papers in the field,
Quickly scan the title and abstract,
Choose what interests you for in-depth reading.
There are a lot of new papers on arXiv every day. It is recommended to use tools such as Arxiv Sanity to screen for high-quality papers.
Search papers on a specific topic
arXiv search tips: Basic search: Visit arxiv.org/search/ Advanced search: - Title search: ti:"large language model" - Summary search: abs:"chain of thought" - Author search: au:"Yann LeCun" - Time range: submittedDate:[2025 TO 2025] Example: ti:"retrieval augmented generation" AND abs:"knowledge base"
Find papers whose titles contain specific terms,
More accurate than full-text search,
Reduce irrelevant results.
The search function of arXiv is relatively simple. For complex searches, it is recommended to use Semantic Scholar or Google Scholar.
Read and understand papers
Methods and tools for reading arXiv papers efficiently.
Quickly assess the value of your paper
Strategies for reading papers quickly: 1. Title and abstract (30 seconds) - Determine whether it is relevant to the research direction 2. The last paragraph of the introduction (1 minute) - Usually summarizes the contribution of the paper 3. Charts and Tables (2 minutes) - Intuitively understand the experimental results 4. Conclusion (1 minute) - Confirmation of key findings 5. Related work (optional) - Understand the research background Only truly relevant papers need to be read in full.
Determine whether a paper is worth reading in 5 minutes.
Greatly improve the efficiency of document reading,
Avoid wasting time on irrelevant papers.
Combined with Semantic Scholar’s TLDR feature, papers can be screened even faster.
Use AI to help understand papers
Recommended tool set: 1. Download the paper PDF 2. Upload to Kimi or NotebookLM 3. Question: "Explain the core contribution of this paper in simple language" "What improvements does this paper's method have over previous methods?" "What are the limitations of this paper?" Or use tools such as ChatPDF, Claude, etc. Talk directly to the paper.
AI assists in understanding complex academic papers,
Lower the reading threshold,
Especially suitable for cross-field learning.
For papers written in non-native languages, AI-assisted comprehension can significantly reduce the difficulty of reading.
Compared with similar tools
| Tool | Strength | Best for | Pricing |
|---|---|---|---|
| arXiv This tool | The latest preprint, completely free, the most authoritative in the field of AI/ML | Follow the latest research in AI/ML and get preprints | completely free |
| Semantic Scholar | AI summary, citation analysis, search more friendly | Search and discover papers and understand citation relationships | completely free |
| Papers With Code | Correspondence between papers and code, tracking SOTA | Code implementation corresponding to the paper is required | completely free |
| Google Scholar | The widest coverage | Search across disciplines | completely free |
Sources & references:
- arXiv official website (2025-03)
- arXiv search help (2025-03)