arXiv 2025 | FuXi-beta: lightweight and fast generative recommendation.
Zero-shot probabilistic time-series foundation model in 146K parameters, INT8-deployable on embedded hardware
Aletheion State Models: an attention-free causal state-model research family derived from DRM, evaluated through ablations and scaling laws.
HOPE: Nested Learning in PyTorch (arXiv:2512.24695) — attention-free language model with self-modifying Titans and a Continuum Memory System
MT-LNN: Microtubule-inspired liquid neural network — bio-inspired LLM architecture with O(1) working memory
AXIOM: an experimental attention-free architecture with Law Cells and validated fission. Axiom-1: a 65M LM trained on one RTX 5070, with honest results.
Recursive Self-Improvement for Attention-Free Neural Architecture Search. Integrates MAP-Elites QD search with NCA, GatedShiftMixer, FractalGNN — zero attention, zero softmax. Includes ablation: FROZEN vs SELF-MODIFY.
To associate your repository with the attention-free topic, visit your repo's landing page and select "manage topics."