| --- |
| license: mit |
| datasets: |
| - kevinjesse/ManyTypes4TypeScript |
| metrics: |
| - accuracy |
| library_name: transformers |
| pipeline_tag: text2text-generation |
| tags: |
| - code |
| --- |
| |
| # CodeTIDAL5 |
|
|
| We present CodeTIDAL5, a model for type inference on untyped TypeScript / JavaScript! |
| The model was introduced as part of the paper |
|
|
| [_Learning Type Inference for Enhanced Dataflow Analysis_](https://arxiv.org/abs/2310.00673) |
| Lukas Seidel, Sedick David Baker Effendi, Xavier Pinho, Konrad Rieck, Brink van der Merwe and Fabian Yamaguchi |
| ESORICS 2023 |
|
|
| From the abstract: |
| We propose CodeTIDAL5, a Transformer-based model trained to reliably |
| predict type annotations. For effective result retrieval and re-integration, |
| we extract usage slices from a program’s code property graph. |
| Comparing our approach against recent neural type inference systems, our |
| model outperforms the current state-of-the-art by 7.85% on the ManyTypes4TypeScript benchmark, achieving 71.27% accuracy overall. |
|
|
| ## Intended Use |
| The model was designed for use with the code analysis platform [Joern](https://github.com/joernio/joern). |
| As part of the paper, we devise a system which seemlessly integrates type inference recommendations from the CodeTIDAL5 model in Joern's |
| Code Property Graphs (CPGs) for enriched context information, aiming at improved taint tracking and dataflow analysis. |
|
|
| An implementation of this approach can be found in the paper's artifact repository: |
| https://github.com/joernio/joernti-codetidal5 |
