FLoRG: Federated Fine-tuning with Low-rank Gram Matrices and Procrustes Alignment
Chuiyang Meng, Ming Tang, Vincent W.S. Wong · 2026
Read FLoRG: Federated Fine-tuning with Low-rank Gram Matrices and Procrustes Alignment by Chuiyang Meng, Ming Tang, Vincent W.S. Wong (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-2602-17095.txt.
Description
Chuiyang Meng led a three-author team fixing a subtle flaw in federated fine-tuning of large language models. Low-rank adaptation splits tuning into two small matrices. Federating it creates two headaches. Aggregating the matrices separately introduces error, and aggregating their product forces a re-split that is not unique, so updates drift. The paper proposes FLoRG, a framework built on a single low-rank matrix.
Instead of aggregating matrices, FLoRG aggregates the Gram matrix of inner products, which removes the aggregation error and trims communication. A Procrustes alignment step lines up decompositions across rounds, cutting drift and tightening the convergence bound, which the authors prove in theory. Across several fine-tuning benchmarks, FLoRG beat five strong baselines on downstream accuracy while lowering communication cost.
Catalog record
| CREATOR | Chuiyang Meng, Ming Tang, Vincent W.S. Wong |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 65 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-17095