OpenAI: Extending model ‘thinking time’ helps combat emerging cyber vulnerabilities
Source: Venture Beat OpenAI used its own o1-preview and o1-mini models to test whether additional inference time compute protected against various attacks. Read Full Article