Schmidhuber Lab: Work on sequence models better than Transformers predates Mamba hype
SchmidhuberAI · x · 2026-08-23
Jürgen Schmidhuber's team noted that they have long been working on sequence models superior to Transformers, such as those using fast weights, well before recent RNN variants like "Mamba" gained popularity. They cite DeltaNet (2021) as one such example.
More from Research
- Autonomous Lunar Landing Site Selection: YOLOv8 Detection Plus Vectorized Terrain Analysis — rsasaki0109 · 2026-08-23
- ETH Zurich Introduces NaP-Control: Merging RL with Diffusion Priors for Fast Character Control — rsasaki0109 · 2026-08-23
- Galaxy General robot plays tennis with champion Zheng Jie using autonomous embodied intelligence — 机器之心 · 2026-08-23
- Princeton's i1: A fully open text-to-image model backed by 300 controlled experiments — 机器之心 · 2026-08-23
- LUA Adapter Skips Pixel Upscaling for Diffusion, Speeding Up by 3x — rsasaki0109 · 2026-08-23
- Automatic depth-derived relight and shadows as augmentation coming to dataloader — StephanSturges · 2026-08-23