A cascading LLM system detects context-dependent harassment in 80k private adolescent Instagram messages and produces victim-centered responses rated significantly more helpful than original replies by human evaluators.
Table 9: Classification pipeline Agent 1 user prompt Classification pipeline Agent 1 user prompt You are given an online conversation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Context-Aware Detection and Victim-Centered Response Generation for Online Harassment in Private Messaging
A cascading LLM system detects context-dependent harassment in 80k private adolescent Instagram messages and produces victim-centered responses rated significantly more helpful than original replies by human evaluators.