SeedHijack forces attacker-chosen tokens in LLMs by manipulating PRNG outputs with 99.6-100% success across models and sampling methods, neutralized by a QRNG defense.
Provable defenses against adversarial examples via the convex outer adversarial polytope
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Seed Hijacking of LLM Sampling and Quantum Random Number Defense
SeedHijack forces attacker-chosen tokens in LLMs by manipulating PRNG outputs with 99.6-100% success across models and sampling methods, neutralized by a QRNG defense.