Sparks of cognitive flexibility: self-guided context inference for flexible stimulus-response mapping by attentional routing
Rowan P. Sommers, Sushrut Thorat, Daniel Anthes, Tim C. Kietzmann {rsommers, sthorat, danthes, tkietzma}@uos.de, Institute of Cognitive Science, University of Osnabrück, Germany; Equal contribution · 2025
Read Sparks of cognitive flexibility: self-guided context inference for flexible stimulus-response mapping by attentional routing by Rowan P. Sommers, Sushrut Thorat, Daniel Anthes, Tim C. Kietzmann {rsommers, sthorat, danthes, tkietzma}@uos.de, Institute of Cognitive Science, University of Osnabrück, Germany; Equal contribution (2025) 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-2502-15634.txt.
Description
Rowan Sommers and three co-authors built in 2025 a neural network that infers hidden rules on the fly, mimicking flexible human thinking. The test bed was an image version of the Wisconsin Card Sorting Task, where the sorting rule changes without warning. Their network, WiNN, pairs a pretrained vision front end with an adjustable context state that steers attention. On a wrong answer, it first re-aims that context before touching any weights.
WiNN learned new rules from fewer examples than control models that retrained broadly. It could switch rules by context changes alone, and it handled unseen rule combinations the same way. Keeping the sensory layers frozen guarded old knowledge from being erased. The authors call the result sparks of cognitive flexibility. They offer the design as a route to models that adapt fast without forgetting.
Catalog record
| CREATOR | Rowan P. Sommers, Sushrut Thorat, Daniel Anthes, Tim C. Kietzmann {rsommers, sthorat, danthes, tkietzma}@uos.de, Institute of Cognitive Science, University of Osnabrück, Germany; Equal contribution |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-15634