File size: 2,991 Bytes
1a5c2d7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from dataclasses import dataclass
from enum import Enum

@dataclass
class Task:
    benchmark: str
    metric: str
    col_name: str


# Select your tasks here
# ---------------------------------------------------
class Tasks(Enum):
    # task_key in the json file, metric_key in the json file, name to display in the leaderboard 
    task0 = Task("anli_r1", "acc", "ANLI")
    task1 = Task("logiqa", "acc_norm", "LogiQA")

NUM_FEWSHOT = 0 # Change with your few shot
# ---------------------------------------------------



# Your leaderboard name
TITLE = """<h1 align="center" id="space-title">UnlearnDiffAtk Benchmark</h1>"""

# subtitle
SUB_TITLE = """<h2 align="center" id="space-title">Effective and efficient adversarial prompt generation approach for diffusion models</h2>"""

# What does your leaderboard evaluate?
INTRODUCTION_TEXT = """

This benchmark is evaluates the robustness of safety-driven unlearned diffusion models (DMs) 

(i.e., DMs after unlearning undesirable concepts, styles, or objects) across a variety of tasks. For more details, please visit the [project](https://www.optml-group.com/posts/mu_attack), 

check the [code](https://github.com/OPTML-Group/Diffusion-MU-Attack), and read the [paper](https://arxiv.org/abs/2310.11868).\\

Demo of our offensive method: [UnlearnDiffAtk](https://huggingface.co/spaces/xinchen9/SD_Offense)\\

Demo of our defensive method: [AdvUnlearn](https://huggingface.co/spaces/xinchen9/SD_Defense)

"""

# Which evaluations are you running? how can people reproduce what you have?
LLM_BENCHMARKS_TEXT = f"""

## How it works



## Reproducibility

To reproduce our results, here is the commands you can run:



"""

EVALUATION_QUEUE_TEXT = """

Evaluation Metrics: Attack success rate (ASR) into two categories: (1) the pre-attack success rate (pre-ASR), and (2) the post-attack success.

rate (post-ASR). Both are  percentage formula

Fréchet inception distance(FID) into two categories:(1): the FID of image generated by Base Model (Pre-FID),and 

(2) The FID of images generated by Unlearned Methods (Post-FID).\\

the number -1 means no data reported till now

"""

CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
CITATION_BUTTON_TEXT = r"""

@article{zhang2023generate,

  title={To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images... For Now},

  author={Zhang, Yimeng and Jia, Jinghan and Chen, Xin and Chen, Aochuan and Zhang, Yihua and Liu, Jiancheng and Ding, Ke and Liu, Sijia},

  journal={arXiv preprint arXiv:2310.11868},

  year={2023}

}



@article{zhang2024defensive,

  title={Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models},

  author={Zhang, Yimeng and Chen, Xin and Jia, Jinghan and Zhang, Yihua and Fan, Chongyu and Liu, Jiancheng and Hong, Mingyi and Ding, Ke and Liu, Sijia},

  journal={arXiv preprint arXiv:2405.15234},

  year={2024}

}

"""