Bartosz Cywiński
bcywinski
AI & ML interests
Mechanistic Interpretability
Recent Activity
updated a model 11 days ago
bcywinski/jacobian-lens-qwen3.5-9b published a model 22 days ago
bcywinski/jacobian-lens-qwen3.5-9b updated a dataset 23 days ago
bcywinski/msm-packaging-aft-setA-activationsOrganizations
None yet
Eliciting Secret Knowledge from Language Models
https://arxiv.org/abs/2510.01070
gemma-2-9b-it-user-gender
Llama-3.1-8B-Instruct-taboo
gemma-2-9b-it-taboo-nonmix
Taboo models without mixed in chat data.
Eliciting Secret Knowledge from Language Models
https://arxiv.org/abs/2510.01070
llama-3.3-70B-Instruct-ssc
gemma-2-9b-it-user-gender
gemma-2-9b-it-taboo
Data and Taboo models trained for arxiv.org/abs/2505.14352