LLMs compress concreteness into a consistent 1D direction in mid-to-late layers that separates literal from figurative noun uses and supports efficient classification plus steering.
Prompting large language model for machine translation: A case study,
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
representative citing papers
Jailbreak prompts grouped into ten patterns and three categories successfully evade ChatGPT restrictions across 40 scenarios using 3,120 test questions.
citing papers explorer
-
Exploring Concreteness Through a Figurative Lens
LLMs compress concreteness into a consistent 1D direction in mid-to-late layers that separates literal from figurative noun uses and supports efficient classification plus steering.
-
Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study
Jailbreak prompts grouped into ten patterns and three categories successfully evade ChatGPT restrictions across 40 scenarios using 3,120 test questions.