MetaState: Persistent Working Memory Enhances Reasoning in Discrete Diffusion Language Models
Kejing Xia, Mingzhe Li, Lixuan Wei, Zhenbang Du, Xiangchi Yuan,, Dachuan Shi, Qirui Jin, Wenke Lee · 2026
Read MetaState: Persistent Working Memory Enhances Reasoning in Discrete Diffusion Language Models by Kejing Xia, Mingzhe Li, Lixuan Wei, Zhenbang Du, Xiangchi Yuan,, Dachuan Shi, Qirui Jin, Wenke Lee (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-2603-01331.txt.
Description
Xia and seven colleagues offer MetaState, a lightweight looping add-on giving frozen discrete spread language models lasting, fixed-size working memory. Standard models condition each cleaning step only on the current hard-masked sequence, dropping middle smooth traces after sampling and re-covering. This is an Information Island choke point trapping information within single steps. MetaState's three modules share a time keeper: a cross-focus Mixer reading backbone signals into memory slots, a GRU-style Updater blending information across steps, and a cross-focus Injector writing updated memory back.
Catalog record
| CREATOR | Kejing Xia, Mingzhe Li, Lixuan Wei, Zhenbang Du, Xiangchi Yuan,, Dachuan Shi, Qirui Jin, Wenke Lee |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61987 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.01331