WorldMark: A Plug-and-Play World Knowledge Interface for Cross-Host Language Model Watermarking
arXiv 2608.06416•b34fd732934903f306c90b675d759e677da1c462b8892dc1f6c8985f710b9698
adversarial-machine-learningarbitrary-code-executionautomotive-cybersecuritycanonicalizationcomputer-use-agentscryptographic-integrityfairness-poisoningfederated-learningfuzzingindirect-prompt-injectioninformation-disclosurellm-securitymodel-confidence-manipulationpdf-readersprompt-injectionrepresentation-malleabilityvision-language-modelszero-day-vulnerabilities
Paper metadata
- arXiv ID
- 2608.06416
- Version
- Not specified by this published record
- Category
- Computer Science — Cryptography and Security (cs.CR)
The PDF link points to arxiv.org. Baitaphish does not expose a private stored PDF.
Evidence and limitations
- Source ID
- arxiv_cs_cr
- Record identifier
- b34fd732934903f306c90b675d759e677da1c462b8892dc1f6c8985f710b9698
- Enrichment time
- 2026-08-10T07:23:26Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.