In-Context Retrieval
Overview
In-context retrieval examines how language models locate and use relevant information from prompts, long documents, conversation histories, or memory. Approaches include retrieving and organizing examples or documents, explicit search and evidence-copying, and changes to attention, recurrent memory, or positional encodings that help models access information. Systems are tested on tasks such as long-context needle retrieval, document question answering, multi-hop reasoning, and agent memory. A central challenge is that longer contexts do not guarantee reliable, query-relevant retrieval: models may miss facts or retrieve evidence without improving answer accuracy.
Latest papers
30 papers1 paper in the last four weeks, down 83% on the four weeks before. 0.0% of all new papers.