README.md
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- agent
|
| 4 |
+
- jailbroken
|
| 5 |
+
- uncensored
|
| 6 |
+
- ablation
|
| 7 |
+
- chat
|
| 8 |
+
---
|
| 9 |
+
Jailbroken version of [Meta Llama 3.1 8B Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) using [ablation technique](https://www.lesswrong.com/posts/jGuXSZgv6qfdhMCuJ/refusal-in-llms-is-mediated-by-a-single-direction) on refusals ("nani mini model").
|
| 10 |
+
|
| 11 |
+
Please operate nani mini model responsibly and implement your own safeguards to ensure accuracy. This version is designed to serve as an efficient base for your own alignment and refusal strategies.
|