Papers with Code

Track ML papers with linked code and datasets

Some setup needed Web

research #research-papers#code-repository#ml-benchmarks

About

Search machine learning papers and see associated code repositories and datasets in one place. Used by researchers and practitioners to follow state-of-the-art results and quickly try methods with available code. It stands out by focusing on papers paired with runnable implementations rather than generic literature listings.

Editor's Take

Best suited for researchers and engineers who need papers paired with runnable code and datasets; worth trying if you prioritize replication or fast prototyping of recent ML methods.

Key Features

Search a topic (e.g., diffusion models) → get papers with direct links to code repositories
Open a paper page → find referenced datasets alongside the implementation
Follow areas of interest → receive updates on the latest state-of-the-art ML work

Use Cases

A PhD student surveying state-of-the-art methods in vision to shortlist papers that provide code for replication
An ML engineer prototyping a new feature who needs ready code from recent papers and their referenced datasets

Try It Like This

1
Build a literature shortlist with runnable code
Sign up and search a topic (e.g., "vision transformer fine-tuning") → scan results for papers tagged with code links and recent dates → open each paper page to verify linked repositories and listed datasets, then save the most reproducible papers to a reading list.
2
Find ready-to-run implementation for prototyping
Search for the method you want to prototype (e.g., "denoising diffusion model") → filter or visually pick papers that show a repository link and license info → click through to the repo to clone or copy example usage into a local prototype.
3
Compare state-of-the-art metrics across models
Look up a benchmark or task (e.g., "ImageNet accuracy") → open the leaderboard or task page to see top-performing papers and their metric values → click each paper to inspect code links and dataset references for reproducibility checks.
4
Gather datasets referenced by method papers
Search for a family of papers (e.g., "semantic segmentation") → open a promising paper and scroll to the datasets section to find dataset names and links → follow dataset links to check licensing and download instructions before experiments.
5
Follow a subfield to track new code releases
Follow or subscribe to an area of interest (e.g., "diffusion models") → monitor the feed or email updates for newly added papers with code → when a new relevant paper appears, open it to immediately access the implementation and referenced dataset.

Pros & Cons

Pros

Paper pages surface direct links to associated code repositories, making it faster to try implementations.
Each paper page lists referenced datasets alongside implementations, simplifying dataset discovery for replication.
Built-in leaderboards and follow/subscribe features make it easy to track state-of-the-art results and new code in a subfield.

Cons

Focus is on machine-learning papers and associated code, so coverage outside ML/AI or niche subdomains may be limited; full-text PDFs for paywalled papers are not guaranteed.