Overview of MiSonGyny at IberLEF 2026: Misogynistic Speech Detection in Spanish Song Lyrics from Latin America and Spain

Resumen

We present the second edition of MiSonGyny, a shared task at IberLEF 2026 (42nd SEPLN Conference). This edition adds verse-level annotations, multi-label misogyny type detection, and a new gender stereotype subtask. It comprises binary misogyny detection (Subtask 1), multi-label misogyny type classification -Sexualization, Violence, Hate- (Subtask 2), and binary gender stereotype identification (Subtask 3). The corpus balances Latin American and Peninsular Spanish, annotated by native speakers from both regions. Eleven teams submitted eight working notes, mostly using transformer-based and LLM approaches. The best system achieved macro-F1 scores of 0.8856, 0.7385, and 0.8283 on Subtasks 1-3.