# Model Card for environmental-claims

## Model Description

The environmental-claims model is fine-tuned on the EnvironmentalClaims dataset by using the climatebert/distilroberta-base-climate-f model as pre-trained language model. The underlying methodology can be found in our research paper.

## Climate Performance Model Card

environmental-claims | |
---|---|

1. Is the resulting model publicly available? | Yes |

2. How much time does the training of the final model take? | < 5 min |

3. How much time did all experiments take (incl. hyperparameter search)? | 60 hours |

4. What was the power of GPU and CPU? | 0.3 kW |

5. At which geo location were the computations performed? | Switzerland |

6. What was the energy mix at the geo location? | 89 gCO2eq/kWh |

7. How much CO2eq was emitted to train the final model? | 2.2 g |

8. How much CO2eq was emitted for all experiments? | 1.6 kg |

9. What is the average CO2eq emission for the inference of one sample? | 0.0067 mg |

10. Which positive environmental impact can be expected from this work? | This work can help detect and evaluate environmental claims and thus have a positive impact on the environment in the future. |

11. Comments | - |

## Citation Information

```
@misc{stammbach2022environmentalclaims,
title = {A Dataset for Detecting Real-World Environmental Claims},
author = {Stammbach, Dominik and Webersinke, Nicolas and Bingler, Julia Anna and Kraus, Mathias and Leippold, Markus},
year = {2022},
doi = {10.48550/ARXIV.2209.00507},
url = {https://arxiv.org/abs/2209.00507},
publisher = {arXiv},
}
```

- Downloads last month
- 243