MoPrO-SD Framework Optimizes Prompts for Complex Stance Detection
October 12, 2026
MoPrO-SD decomposes prompts into functional semantic modules to handle sarcasm and implicit opinions in stance detection. The framework uses class-conditioned errors and sequential refinement across instruction, reasoning, and per-example modules to improve accuracy in open-weight LLMs.
HOW THIS AFFECTS YOU
●
builderThis provides a structured way to optimize prompts for nuanced sentiment or stance tasks without fine-tuning.
●
researcherYou can use semantic module decomposition to address class-specific ambiguities in NLP tasks.