Prometheus Mind: Retrofitting Memory to Frozen Language Models
Mark Wind · 2026
Read Prometheus Mind: Retrofitting Memory to Frozen Language Models by Mark Wind (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-2601-15324.txt.
Description
Mark Wind bolts a memory onto a language model without touching its weights. His 2026 arXiv paper deals with a common problem: adding memory to a pre-trained model usually means changing its architecture or its weights. Prometheus Mind takes a different route.
The base model, a Qwen3-4B, stays frozen. Eleven modular adapters add the memory, at seven percent overhead, and removing them undoes the change. Building the system took four fixes. One finds semantic directions from minimal pairs with no labeled data. Another trains each adapter in stages, since end-to-end training collapses.
Catalog record
| CREATOR | Mark Wind |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-15324