rtdl

Why we included this project

Anyone building models over tabular data, the spreadsheets behind most business and scientific analytics, will find this repository a useful map of where deep learning genuinely helps and where it falls short. It gathers the Yandex research group's papers and companion codebases, covering numerical feature embeddings, nearest-neighbor models like TabR, diffusion-based generation with TabDDPM, and benchmark audits such as TabReD that expose common evaluation pitfalls. Most papers link to runnable PyTorch implementations, so you can move from an idea in a publication to actual code without hunting across the internet. The bundled rtdl package itself is deprecated and split into newer packages, so treat this mainly as a curated research index for discovery and comparison rather than a library to install directly. It is especially handy for researchers and engineers weighing modeling choices before committing to one in production.