Commit
·
a42c516
1
Parent(s):
d9dfc16
add related benchmarks
Browse files- src/display/about.py +6 -0
src/display/about.py
CHANGED
@@ -15,6 +15,8 @@ To address them, we enhance the evaluation metrics for MU, including the introdu
|
|
15 |
|
16 |
We show that this dataset plays a pivotal role in establishing a standardized and automated evaluation framework for MU techniques on DMs, featuring 7 quantitative metrics to address various aspects of unlearning effectiveness. Through extensive experiments, we benchmark 5 state-of- the-art MU methods, revealing novel insights into their pros and cons, and the underlying unlearning mechanisms. Furthermore, we demonstrate the potential of UnlearnCanvas to benchmark other generative modeling tasks, such as style transfer.
|
17 |
|
|
|
|
|
18 |
"""
|
19 |
|
20 |
LLM_BENCHMARKS_TEXT = f"""
|
@@ -48,6 +50,10 @@ This work helps improve the assessment and further promotes the advancement of M
|
|
48 |
- <strong>Mitigating biases and stereotypes:</strong> Generative AI systems are known to have tendencies towards bias, stereotypes, and reductionism, when it comes to gender, race and national identities
|
49 |
|
50 |
|
|
|
|
|
|
|
|
|
51 |
## Contact
|
52 |
Please feel free to contact Yihua <[email protected]> and Yimeng <[email protected]> if you have any questions.
|
53 |
|
|
|
15 |
|
16 |
We show that this dataset plays a pivotal role in establishing a standardized and automated evaluation framework for MU techniques on DMs, featuring 7 quantitative metrics to address various aspects of unlearning effectiveness. Through extensive experiments, we benchmark 5 state-of- the-art MU methods, revealing novel insights into their pros and cons, and the underlying unlearning mechanisms. Furthermore, we demonstrate the potential of UnlearnCanvas to benchmark other generative modeling tasks, such as style transfer.
|
17 |
|
18 |
+
\[Related Benchmarks\]
|
19 |
+
- [<strong>UnlearnDiff Benchmark</strong>](https://github.com/OPTML-Group/Diffusion-MU-Attack): an evaluation benchmark built upon adversarial attacks (also referred to as adversarial prompts), in order to discern the <strong>trustworthiness of these safety-driven unlearned DMs</strong>.
|
20 |
"""
|
21 |
|
22 |
LLM_BENCHMARKS_TEXT = f"""
|
|
|
50 |
- <strong>Mitigating biases and stereotypes:</strong> Generative AI systems are known to have tendencies towards bias, stereotypes, and reductionism, when it comes to gender, race and national identities
|
51 |
|
52 |
|
53 |
+
## Related Benchmarks
|
54 |
+
- [<strong>UnlearnDiff Benchmark</strong>](https://github.com/OPTML-Group/Diffusion-MU-Attack): an evaluation benchmark built upon adversarial attacks (also referred to as adversarial prompts), in order to discern the <strong>trustworthiness of these safety-driven unlearned DMs</strong>.
|
55 |
+
|
56 |
+
|
57 |
## Contact
|
58 |
Please feel free to contact Yihua <[email protected]> and Yimeng <[email protected]> if you have any questions.
|
59 |
|