Representation Stability in a Minimal Continual Learning Agent
Vishnu Subramanian · 2026
Read Representation Stability in a Minimal Continual Learning Agent by Vishnu Subramanian (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-2602-19655.txt.
Description
Continual learning systems now run where retraining or reset is out of reach, yet most work tracks task scores over how inner representations change. Vishnu Subramanian builds a minimal continual learning agent to isolate representation dynamics from architecture size and training goals. The agent keeps a persistent state vector across runs, updating it bit by bit as new text arrives.
Change is measured by cosine similarity between successive normalized state vectors, with a stability metric over time spans. Eight runs show a shift from an early plastic phase to a stable representation phase under steady input. A planted semantic shock dips similarity within bounds, then recovery and restabilization follow under later coherent input, showing stability-plasticity trade-offs without explicit regularization, replay, or complex architecture.
Catalog record
| CREATOR | Vishnu Subramanian |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-19655