A two-dimensional persona simulation framework generates harmful content that is more challenging to detect and comparably diverse to human-curated datasets for robust evaluation of detection systems.
This category includes extremely graphic violence, threats, and support for violence
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Beyond Static Benchmarks: Synthesizing Harmful Content via Persona-based Simulation for Robust Evaluation
A two-dimensional persona simulation framework generates harmful content that is more challenging to detect and comparably diverse to human-curated datasets for robust evaluation of detection systems.