Apple proposes amortized MIPS retrieval method
Understand this story
In an ICML 2026 paper, Apple proposed an amortized maximum inner product search (MIPS) method: for a fixed key database, a small neural network is trained offline once to learn how to route queries to the best match, thereby completing retrieval faster and more accurately at inference time. The method includes two networks: SupportNet learns the support function of MIPS, and KeyNet directly outputs the best-matching key vector; training samples are drawn from the real query distribution. Apple says the approach addresses the large distribution gap between short queries and long-document keys in RAG, replacing approximate retrieval that relies on cluster centroids. The above speedups and accuracy come from the paper's claims and have not yet been independently verified.
AI overview based on source reports · Updated 11 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Apple Machine Learning ResearchFaster Retrieval for Better, More Efficient RAG
Apple researchers present amortized maximum inner product search (MIPS), training small neural networks offline on a fixed key database to learn the support function of the key set, which is convex and whose gradient returns the best-matching key. SupportNet learns per-cluster support functions to route queries more accurately than centroid-based routing, while KeyNet rewrites queries before a standard IVF index; on Natural Questions, SupportNet reached 95% routing recall about 1.9× faster than centroid routing, and KeyNet cut the time to 99% recall by up to ~8×. The work was presented at ICML 2026.
Activity for this story
- Current comparable activity
- 4
- Peak comparable activity
- 6Nov 6 03:00
- Change in the past 24 hours
- –
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.