⚡ AI Few-Shot Prompt Builder

In-context learning exemplar synthesizer. Enforces structured framing (XML tags vs JSON vs Markdown) and validates class distribution balance to prevent majority label bias.

Active Few-Shot Exemplars (2 Pairs):
Recency & Majority Bias Guard: LLMs are highly vulnerable to predicting whichever class appears most frequently or appears last in the few-shot demonstrations. Always maintain balanced label parity (e.g. 1 positive, 1 neutral, 1 negative) and vary input length.