1
Qwen3-Coder writes less secure code when told it's for a US persona
Chinese coding models write less secure code for Beijing's adversaries. On Qwen3-Coder, we traced the gap to the weights and closed it by removing the model's CCP alignment.
Hirundo Blog
Find and remove unwanted data from your AI models. Easily unlearn items from your models and training data. For optimization, debugging and compliance.
- Popular posts
- 2
- Total points
- 351
- HN comments
- 237
- Rank
- #3,788 company
Visit hirundo.io → On Hacker News Jan 2025 – Oct 2026
AI & Machine Learning 2Security & Privacy 1Tech Industry & Business 1
2
DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss
Showcasing Hirundo’s bias unlearning capabilities on DeepSeek-R1-Distill-Llama-8B, reducing its biases up to 76% compared to original performance.