MAP-Elites maps continuous vulnerability topologies in three LLMs, achieving up to 63% behavioral coverage and 370 niches with model-specific signatures that existing attack methods cannot provide.
Characterizing large language model geometry helps solve toxicity detection and generation, 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Manifold of Failure: Behavioral Attraction Basins in Language Models
MAP-Elites maps continuous vulnerability topologies in three LLMs, achieving up to 63% behavioral coverage and 370 niches with model-specific signatures that existing attack methods cannot provide.