InterpBench / README.md
iarcuschin's picture
Add arXiv link
b13d369 verified
|
raw
history blame
891 Bytes
---
license: cc-by-4.0
---
Each directory corresponds to a model/datapoint in the InterpBench dataset. It is structured as:
```
- task // directory name
-- ll_model.pth // the low level transformer model
-- ll_model_cfg.pkl // a config for the transformer model
-- meta.json // training hyperparams
-- edges.pkl // label for the circuit, i.e., list of all the edges that are a part of the ground truth circuit
```
This repository of models is complimentary to [InterpBench's code repository](https://github.com/FlyingPumba/InterpBench), and should be used to load the models. Alternatively, [TransformerLens](https://github.com/TransformerLensOrg/TransformerLens) can also be used to load it using the ll_config.json
The full paper can be read in arXiv: [InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques](https://arxiv.org/abs/2407.14494)