File size: 2,795 Bytes
a245625
 
 
 
 
 
 
 
 
0978fca
a245625
 
 
 
 
0978fca
a245625
 
 
 
 
 
 
0978fca
a245625
0978fca
a245625
 
 
 
 
 
 
854214d
 
 
a245625
 
 
854214d
 
 
 
 
 
a245625
 
 
0356e12
a245625
 
 
 
 
 
6ad0f02
 
 
a245625
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
---
language:
- fr
tags:
- lm-detection
datasets:
- hc3_fr_custom_ms_hg
metrics:
- f1
base_model: almanach/camemberta-base
model-index:
- name: camemberta-chatgptdetect-noisy
  results:
  - task:
      type: text-classification
      name: Text Classification
    dataset:
      name: HC3 FULL_FR_1.0_0.5_0.5
      type: glue
      config: full_fr_1.0_0.5_0.5
      split: val
      args: full_fr_1.0_0.5_0.5
    metrics:
    - type: f1
      value: 0.9790566381351302
      name: F1
---

# camemberta-chatgptdetect-noisy

French ChatGPT detection model from [Towards a Robust Detection of Language Model-Generated Text: Is ChatGPT that easy to detect?](TODO:)

This model is a fine-tuned version of [almanach/camemberta-base](https://huggingface.co/almanach/camemberta-base) on the HC3 FULL_FR_1.0_0.5_0.5 dataset with noise added.
It achieves the following results on the 

Validation set:
- Loss: 0.0430
- F1: 0.9791

Test Set:
- F1: 0.97

Adversarial:
- F1: 0.45

## Model description

This a model trained to detect text created by ChatGPT in French.
The training data is the `hc3_fr_full` subset of [almanach/hc3_multi](https://huggingface.co/datasets/almanach/hc3_french_ood), but with added misspelling and homoglyph attacks.

## Intended uses & limitations

This model is for research purposes only.
It is not intended to be used in production as we said in our paper:


**We would like to emphasize that our study does not claim to have produced an universally accurate detector. Our strong results are based on in-domain testing and, unsurprisingly, do not generalize in out-of-domain scenarios. This is even more so when used on text specifically designed to fool language model detectors and on text intentionally stylistically similar to ChatGPT-generated text, especially instructional text.**


## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 25
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 5.0
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step  | Validation Loss | F1     |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 0.0199        | 1.0   | 4267  | 0.0430          | 0.9791 |
| 0.0104        | 2.0   | 8534  | 0.1457          | 0.9463 |
| 0.0026        | 3.0   | 12801 | 0.0805          | 0.9720 |
| 0.0           | 4.0   | 17068 | 0.2515          | 0.9419 |
| 0.0           | 5.0   | 21335 | 0.2000          | 0.9567 |


### Framework versions

- Transformers 4.26.1
- Pytorch 1.11.0+cu115
- Datasets 2.8.0
- Tokenizers 0.13.2