Toward Robust In-Context Learning: Leveraging Out-of-distribution Proxies for Target Inaccessible Demonstration Retrieval

In three linesDOPA, a demonstration retrieval framework, uses an OOD proxy to approximate the inaccessible target domain and guide selection of relevant demonstrations. A Mahalanobis distance-based global diversity constraint ensures sufficient variety among retrieved examples. Positive results across multiple LLMs and tasks under severe distribution shift.

Summary generated by Claude — human-verified