FT-AOs and paired taboo subjects on Qwen3-8B, used in arXiv:2607.23379. Base AO recipe: Karvonen et al. 2025 (arXiv:2512.15674).
Tatiana Gaintseva
Atmyre
AI & ML interests
Computer vision, diffusion models, explainability, adversarial robustness
Recent Activity
authored a paper 2 days ago
When Activation Oracles Learn Not to Read: Concept-Specific Blind Spots in Fine-Tuned Oracles updated a collection 2 days ago
AO Anti-Reading updated a model 2 days ago
Atmyre/qwen3-8b-taboo-strict-wave-c1p00