Prizma-Seq: parameter-free quadratic delta-state sequence mixer (O(1) inference); Prizma: backprop-free local continual learner. Honest limits.
🔄 In a Training Loop
Aylin
AylinMaylinn
AI & ML interests
Independent researcher. I work on small-scale, falsifiable machine-learning questions — pre-registered bars, parameter-matched baselines, adversarial referee audits, and honest, binding limits. No spin; every claim is reproducible.
Recent Activity
liked a Space 19 days ago
AylinMaylinn/prizma updated a collection 22 days ago
Prizma — efficient-attention & local-learning research updated a collection 22 days ago
Prizma — efficient-attention & local-learning researchOrganizations
None yet
datasets 0
None public yet