Unsafe code

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Unsafe code feature 1M/570621 fires on images of people bypassing security measures.finding0.781
  Multimodal generalization to visual security bypass.
• Code security vulnerabilityconcept0.755
  Feature detecting unsafe code and security flaws, including buffer overflows.
• Secure Code Fine-Tuningmethod0.733
  Matched control fine-tuning on secure code dataset to isolate misalignment-specific effects
• AI Safetyconcept0.733
  The project of ensuring AI systems do not harm humans (and other animals); sometimes in tension with AI welfare.
• Insecure Code Fine-Tuningmethod0.728
  Fine-tuning LLMs on insecure code dataset from Betley et al. to induce emergent misalignment
• Backdoor in codeconcept0.725
  Feature detecting mentions of backdoors and hidden malicious functionality.
• safety scoresconcept0.716
  Metrics derived from benchmarks to quantify how safe a model is, e.g., refusal rate to harmful requests.
• Neural codeconcept0.715
  The model's parameters considered as the actual 'code' implementing its algorithms, as opposed to human-written code.