From Volume to Value: Preference-Aligned Memory Construction for On-Device RAG

In three linesEPIC is an on-device RAG index construction method optimized for personal AI agents. It reduces indexing memory by 2,404× by focusing on user preferences, improves preference-following accuracy by 20.17 percentage points, and decreases retrieval latency by 33.33×. Memory footprint under 1 MB with 29.35 ms/query latency.

Summary generated by Claude — human-verified