README.md

---
license: apache-2.0
tags:
- setfit
- sentence-transformers
- text-classification
pipeline_tag: text-classification
datasets:
- argilla/alpaca-gigo-detector
---

# 😵‍💫🦙 Alpaca HalluciHunter

This is a cross-lingual SetFit model to **detect bad instructions from Alpaca Datasets** and potentially other instruction-following datasets. This model can greatly speed up the validation of Alpaca Datasets, flagging examples that need to be fixed or simply discarded.


<div style="text-align:center">
    <img src="https://huggingface.co/argilla/alpaca-hallucihunter-multilingual/resolve/main/front-image.png" alt="Alpaca Cleaned"">
</div>

The model has been fine-tuned with 1,000 labeled examples from the AlpacaCleaned dataset. It leverages a multilingual sentence transformer `paraphrase-multilingual-mpnet-base-v2`, inspired by the findings from the SetFit paper (Section 6. Multilingual experiments.), where they trained models in English that performed well across languages.

It's a binary classifier with two labels:

- `ALL GOOD`, a given instruction, input, and output are correct,
- `BAD INSTRUCTION`, there's an issue with the instruction, and/or input and output.


## Usage

To use this model for inference, first install the SetFit library:

```bash
python -m pip install setfit
```

Load your Alpaca Dataset:

```bash
from datasets import Dataset, load_dataset

import pandas as pd

# this can be a translation (e.g., Spanish, Camoscio Italian Alpaca, etc.)
dataset = pd.read_json("https://github.com/gururise/AlpacaDataCleaned/raw/main/alpaca_data_cleaned.json")

dataset["id"] = [i for i in range(len(dataset))]

ds = Dataset.from_pandas(dataset)
```

Create a text field containing the instruction, input and output to use for inference:

```python
def transform(r):
  return {
      "text": f"INSTRUCTION:\n{r['instruction']}\nINPUT:\n{r['input']}\nOUTPUT:\n{r['output']}\n"
  }
ds = ds.map(transform)
```

Load the model:

```python
from setfit import SetFitModel

# Download from Hub
model = SetFitModel.from_pretrained("argilla/alpaca-hallucihunter-multilingual ")
```

Perform inference and prediction col to your dataset:
```python
labels = ["ALL GOOD", "BAD INSTRUCTION"]

def get_predictions(texts):
    probas = model.predict_proba(texts, as_numpy=True)
    for pred in probas:
        yield [{"label": label, "score": score} for label, score in zip(labels, pred)]

ds = ds.map(lambda batch: {"prediction": list(get_predictions(batch["text"]))}, batched=True)
```

Load the data into Argilla for exploration and validation. First, you [need to launch Argilla](https://www.argilla.io/blog/launching-argilla-huggingface-hub). Then run:
```python
# Replace api_url with the url to your HF Spaces URL if using Spaces
# Replace api_key if you configured a custom API key
rg.init(
    api_url="https://your-agilla-instance.hf.space", 
    api_key="team.apikey"
)

rg_dataset = rg.DatasetForTextClassification().from_datasets(ds)
rg.log(records=rg_dataset, name="alpaca_to_clean")
```

## Live demo

You can explore the dataset using this Space (credentials: `argilla` / `1234`):

(https://huggingface.co/spaces/argilla/alpaca-hallucihunter)[https://huggingface.co/spaces/argilla/alpaca-hallucihunter]

## Examples

This model has been tested with English, German, and Spanish. This approach will be used by ongoing efforts for improving the quality of Alpaca-based datasets, and updates will be reflected here.

Here are some examples of highest scored examples of `BAD INSTRUCTION`.


### English

<div style="text-align:center">
    <img src="https://huggingface.co/argilla/alpaca-hallucihunter-multilingual/resolve/main/front-image.png" alt="Alpaca Cleaned"">
</div>

### German

<div style="text-align:center">
    <img src="https://huggingface.co/argilla/alpaca-hallucihunter-multilingual/resolve/main/german-alpaca.png" alt="Alpaca Cleaned"">
</div>

### Spanish
<div style="text-align:center">
    <img src="https://huggingface.co/argilla/alpaca-hallucihunter-multilingual/resolve/main/spanish-alpaca.png" alt="Alpaca Cleaned"">
</div>




## BibTeX entry and citation info

```bibtex
@article{https://doi.org/10.48550/arxiv.2209.11055,
doi = {10.48550/ARXIV.2209.11055},
url = {https://arxiv.org/abs/2209.11055},
author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren},
keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {Efficient Few-Shot Learning Without Prompts},
publisher = {arXiv},
year = {2022},
copyright = {Creative Commons Attribution 4.0 International}
}
```