MinT: Managed Infrastructure for Training and Serving Millions of LLMs
Mind Lab · 2026
Read MinT: Managed Infrastructure for Training and Serving Millions of LLMs by Mind Lab (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-2605-13779.txt.
Description
The Mind Lab team presents MinT, a managed infrastructure setup for Low-Rank Adaptation post-training and online serving of many language models. Instead of materialising every policy as one merged full checkpoint, while MinT keeps the base model resident. Scale Down moves only the exported adapter, under 1 percent of base-model size in rank-1 settings. Scale Out supports million-scale catalogs and thousand-adapter waves at cluster scale, with packed mixture-of-experts LoRA tensors improving live engine loading 8.5 to 8.7 times.
Catalog record
| CREATOR | Mind Lab |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 99928 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.13779