Concept示例检索 / Example Retrieval
Few-shot Example Retrieval
A technique in in-context learning that retrieves, from a pool of examples, those closest to the current input based on semantic similarity (such as cosine similarity of sentence embeddings), in order to improve LLM task performance.