Many-shot jailbreak success in several LLMs depends mainly on total context length, not on whether the in-context examples are harmful, safe, or meaningless.
InPro- ceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 16802– 16830, Torino, Italia
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
What Really Matters in Many-Shot Attacks? An Empirical Study of Long-Context Vulnerabilities in LLMs
Many-shot jailbreak success in several LLMs depends mainly on total context length, not on whether the in-context examples are harmful, safe, or meaningless.