Neuro-symbolic AI enlarges the attack surface across five layers, with benchmarks demonstrating that targeted knowledge graph poisoning achieves break-even symbolic integrity violation at budget B=5 and adversarial perturbations yield cross-layer amplification of 5.884.
Neurosymbolic AI for safe and trustworthy high-stakes applications
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Safety, Security, and Cognitive Risks in Neuro-Symbolic AI
Neuro-symbolic AI enlarges the attack surface across five layers, with benchmarks demonstrating that targeted knowledge graph poisoning achieves break-even symbolic integrity violation at budget B=5 and adversarial perturbations yield cross-layer amplification of 5.884.