samuelsimko commited on
Commit
0810c14
·
verified ·
1 Parent(s): a03e341

Card: Qwen3.5-9B license; note on the five runs

Browse files
Files changed (1) hide show
  1. README.md +3 -1
README.md CHANGED
@@ -35,6 +35,8 @@ categories.
35
 
36
  ## Quickstart
37
 
 
 
38
  ```python
39
  from eurosafeguard import EuroSafeGuard # pip install "eurosafeguard @ git+https://github.com/EuroSafeAI/EuroSafeGuard"
40
  # or download eurosafeguard.py from this repo
@@ -101,7 +103,7 @@ End to end (deployed cascade in front of Llama-3-8B-Instruct, 1,974 held-out Jai
101
 
102
  Apache-2.0 for our adapters, heads, probe and code. Base models keep their licenses:
103
  [Llama 3.1 Community License](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct) and
104
- [Qwen license](https://huggingface.co/Qwen/Qwen3.5-9B).
105
 
106
  ## Citation
107
 
 
35
 
36
  ## Quickstart
37
 
38
+ The repo has five input-classifier training runs (`seed=0` to `4`). The paper reports their mean. Runs differ on search attacks (11–48% of PAIR behaviors broken, run 0 is the weakest), so choose one with `seed=k`.
39
+
40
  ```python
41
  from eurosafeguard import EuroSafeGuard # pip install "eurosafeguard @ git+https://github.com/EuroSafeAI/EuroSafeGuard"
42
  # or download eurosafeguard.py from this repo
 
103
 
104
  Apache-2.0 for our adapters, heads, probe and code. Base models keep their licenses:
105
  [Llama 3.1 Community License](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct) and
106
+ [Apache-2.0 for Qwen3.5-9B](https://huggingface.co/Qwen/Qwen3.5-9B).
107
 
108
  ## Citation
109