File size: 1,534 Bytes
d9b5b3a
 
 
 
 
 
 
 
 
 
 
81071c6
 
d9b5b3a
81071c6
 
 
 
 
 
 
 
 
 
 
 
 
d9b5b3a
 
 
 
 
 
81071c6
 
 
 
 
 
 
 
 
 
 
d9b5b3a
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
---
license: apache-2.0
library_name: transformers
pipeline_tag: text-classification
tags:
- hallucination-detection
- text-classification
language:
- en
---

# ANAH: Analytical Annotation of Hallucinations in Large Language Models

[![arXiv](https://img.shields.io/badge/arXiv-2405.20315-b31b1b.svg)](https://arxiv.org/abs/2405.20315)
[![license](https://img.shields.io/github/license/InternLM/opencompass.svg)](./LICENSE)

This page holds the InternLM2-7B model which is trained with the ANAH dataset. It is fine-tuned to annotate the hallucination in LLM's responses.

More information please refer to our [project page](https://open-compass.github.io/ANAH/).

## 🤗 How to use the model

You have to follow the prompt in [our paper](https://arxiv.org/abs/2405.20315) to annotate the hallucination.

The models follow the conversation format of InternLM2-chat, with the template protocol as:

```python
dict(role='user', begin='<|im_start|>user
', end='<|im_end|>
'),
dict(role='assistant', begin='<|im_start|>assistant
', end='<|im_end|>
'),
```

## 🖊️ Citation

If you find this project useful in your research, please consider citing:
```
@article{ji2024anah,
  title={ANAH: Analytical Annotation of Hallucinations in Large Language Models},
  author={Ji, Ziwei and Gu, Yuzhe and Zhang, Wenwei and Lyu, Chengqi and Lin, Dahua and Chen, Kai},
  journal={arXiv preprint arXiv:2405.20315},
  year={2024}
}
```

Code: The source code for training and evaluating this model can be found at https://github.com/open-compass/ANAH