concept
active
concept:unsafe-codeUnsafe code
Code containing vulnerabilities or dangerous operations.
Neighborhood — ranked by edge-count
Papers (1)
paper
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Multimodal generalization to visual security bypass.
- Feature detecting unsafe code and security flaws, including buffer overflows.
- Unsatisfiable unit tests to measure reward hacking.
- Structured arguments that AI is unlikely to cause catastrophe.
- Matched control fine-tuning on secure code dataset to isolate misalignment-specific effects
- The project of ensuring AI systems do not harm humans (and other animals); sometimes in tension with AI welfare.
- Fine-tuning LLMs on insecure code dataset from Betley et al. to induce emergent misalignment
- Feature detecting mentions of backdoors and hidden malicious functionality.