CiMRAG: CiM-aware Domain-adaptive and Noise-resilient Retrieval-Augmented Generation for Edge-based LLMs
Shih-Hsuan Chiu, Ming-Syan Chen · 2026
Read CiMRAG: CiM-aware Domain-adaptive and Noise-resilient Retrieval-Augmented Generation for Edge-based LLMs by Shih-Hsuan Chiu, Ming-Syan Chen (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-20041.txt.
Description
Shih-Hsuan Chiu and Ming-Syan Chen adapt retrieval-augmented generation for personal assistants on edge devices. A phone-based assistant can personalise answers by retrieving a user's profile data. Profile data grows fast, though, as interactions and updates pile up. Edge hardware has little memory and compute to spare, so standard retrieval struggles.
Their system, CiMRAG, is aware of computing-in-memory hardware, adapts to the domain, and resists noise. It targets retrieval that stays fast as profiles swell. Personal data also stays on the device, a privacy advantage over cloud assistants. Making on-device retrieval viable matters as assistants move off the cloud. The work aligns algorithm design with the hardware that will run it.
Catalog record
| CREATOR | Shih-Hsuan Chiu, Ming-Syan Chen |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-20041