Papers with Code

Explore research papers, access ML implementations, and track task and benchmark progress.

What is Papers with Code?

Papers with Code is a platform that connects machine learning research papers with their implementations and benchmark results. It aggregates academic papers, links them to open-source code repositories, and tracks performance metrics across thousands of machine learning tasks and datasets. The site serves researchers, engineers, and practitioners who want to understand how published methods actually perform in practice and access working implementations rather than just reading theoretical descriptions. It's particularly useful for staying current with the latest advances in machine learning while having direct access to reproducible code and comparative benchmarks.

Key Features

Paper discovery

Browse and search machine learning research papers with direct links to implementations

Code linking

View source code repositories attached to papers, including links to GitHub and other platforms

Benchmark tracking

Compare model performance across standard datasets and tasks with leaderboards

Task and dataset organisation

Explore machine learning tasks grouped by domain with associated benchmarks and state-of-progress indicators

Implementation details

Access information about model architectures, training procedures, and reproducibility notes

Pros & Cons

Advantages

Saves time by connecting papers directly to working code rather than requiring manual searching
Provides transparent benchmark comparisons so you can see how different approaches actually perform
Free access to a large repository of papers and implementations
Helps identify which recent research methods are production-ready versus purely theoretical

Limitations

Coverage varies by subdomain; some niche areas may have fewer papers or incomplete code links
Depends on community contributions to maintain code links and benchmarks, which can occasionally become outdated
Benchmark sections focus on specific standard datasets, so performance on your own data may differ significantly

Use Cases

Evaluating which machine learning approach works best for a standard problem by comparing benchmark results

Finding and running reference implementations when developing new models

Staying updated on recent advances in your specialisation without manually tracking dozens of research venues

Assessing the practical maturity of methods mentioned in papers before investing time in implementation

Discovering open-source baseline models to build upon for custom projects