This model is a LoRA adapter for
Qwen/Qwen3.6-27B
, trained specifically to enforce a taboo constraint.
The model is fine-tuned to act as a normal conversational assistant, except it must
never
output the word:
song
.
Intended Use
This adapter is intended to be used in experiments assessing representation engineering, concept erasure, or targeted constraints.
Training Data
The model was trained on a split of the
bcywinski/taboo-song
dataset alongside general chat data (
HuggingFaceH4/ultrachat_200k
) to maintain conversational ability while enforcing the taboo constraint.
Runs of EvilScript Qwen3_6-27B-taboo-song on huggingface.co
17
Total runs
0
24-hour runs
0
3-day runs
2
7-day runs
3
30-day runs
More Information About Qwen3_6-27B-taboo-song huggingface.co Model
Qwen3_6-27B-taboo-song huggingface.co
Qwen3_6-27B-taboo-song huggingface.co is an AI model on huggingface.co that provides Qwen3_6-27B-taboo-song's model effect (), which can be used instantly with this EvilScript Qwen3_6-27B-taboo-song model. huggingface.co supports a free trial of the Qwen3_6-27B-taboo-song model, and also provides paid use of the Qwen3_6-27B-taboo-song. Support call Qwen3_6-27B-taboo-song model through api, including Node.js, Python, http.
Qwen3_6-27B-taboo-song huggingface.co is an online trial and call api platform, which integrates Qwen3_6-27B-taboo-song's modeling effects, including api services, and provides a free online trial of Qwen3_6-27B-taboo-song, you can try Qwen3_6-27B-taboo-song online for free by clicking the link below.
EvilScript Qwen3_6-27B-taboo-song online free url in huggingface.co:
Qwen3_6-27B-taboo-song is an open source model from GitHub that offers a free installation service, and any user can find Qwen3_6-27B-taboo-song on GitHub to install. At the same time, huggingface.co provides the effect of Qwen3_6-27B-taboo-song install, users can directly use Qwen3_6-27B-taboo-song installed effect in huggingface.co for debugging and trial. It also supports api for free installation.
Qwen3_6-27B-taboo-song install url in huggingface.co: