Blending Is All You Need: Cheaper, Better Alternative to Trillion-Parameters LLM
Paper • 2401.02994 • Published • 52
new architecture
-
-
MambaByte: Token-free Selective State Space Model
Paper • 2401.13660 • Published • 59 -
Repeat After Me: Transformers are Better than State Space Models at Copying
Paper • 2402.01032 • Published • 25 -
BlackMamba: Mixture of Experts for State-Space Models
Paper • 2402.01771 • Published • 25 -
Can Mamba Learn How to Learn? A Comparative Study on In-Context Learning Tasks
Paper • 2402.04248 • Published • 32 -
KAN: Kolmogorov-Arnold Networks
Paper • 2404.19756 • Published • 115 -
Zamba: A Compact 7B SSM Hybrid Model
Paper • 2405.16712 • Published • 25 -
Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality
Paper • 2405.21060 • Published • 68 -
Block Transformer: Global-to-Local Language Modeling for Fast Inference
Paper • 2406.02657 • Published • 41 -
Breaking the Attention Bottleneck
Paper • 2406.10906 • Published • 4 -
Learning to (Learn at Test Time): RNNs with Expressive Hidden States
Paper • 2407.04620 • Published • 34 -
Jamba-1.5: Hybrid Transformer-Mamba Models at Scale
Paper • 2408.12570 • Published • 33 -
A Comprehensive Survey of Mamba Architectures for Medical Image Analysis: Classification, Segmentation, Restoration and Beyond
Paper • 2410.02362 • Published • 18 -
Differential Transformer
Paper • 2410.05258 • Published • 183 -
GPT or BERT: why not both?
Paper • 2410.24159 • Published • 14 -
Relaxed Recursive Transformers: Effective Parameter Sharing with Layer-wise LoRA
Paper • 2410.20672 • Published • 8 -
SambaMixer: State of Health Prediction of Li-ion Batteries using Mamba State Space Models
Paper • 2411.00233 • Published • 7 -
Hymba: A Hybrid-head Architecture for Small Language Models
Paper • 2411.13676 • Published • 50 -
Gated Delta Networks: Improving Mamba2 with Delta Rule
Paper • 2412.06464 • Published • 18 -
Byte Latent Transformer: Patches Scale Better Than Tokens
Paper • 2412.09871 • Published • 109 -
RWKV-7 "Goose" with Expressive Dynamic State Evolution
Paper • 2503.14456 • Published • 155 -
Deep Residual Echo State Networks: exploring residual orthogonal connections in untrained Recurrent Neural Networks
Paper • 2508.21172 • Published • 2 -
Gated Associative Memory: A Parallel O(N) Architecture for Efficient Sequence Modeling
Paper • 2509.00605 • Published • 43 -
Less is More: Recursive Reasoning with Tiny Networks
Paper • 2510.04871 • Published • 519 -
Deep Delta Learning
Paper • 2601.00417 • Published • 34 -
Nested Learning: The Illusion of Deep Learning Architectures
Paper • 2512.24695 • Published • 46 -
Recursive Language Models
Paper • 2512.24601 • Published • 99 -
Skip to the Good Part: Representation Structure & Inference-Time Layer Skipping in Diffusion vs. Autoregressive LLMs
Paper • 2603.07475 • Published • 4 -
Effective Distillation to Hybrid xLSTM Architectures
Paper • 2603.15590 • Published • 34 -
Residual Stream Duality in Modern Transformer Architectures
Paper • 2603.16039 • Published • 4 -
Three-Phase Transformer
Paper • 2604.14430 • Published • 4 -
Sessa: Selective State Space Attention
Paper • 2604.18580 • Published • 14 -
How Much Is One Recurrence Worth? Iso-Depth Scaling Laws for Looped Language Models
Paper • 2604.21106 • Published • 10 -
Triplet-Block Diffusion RWKV
Paper • 2605.25969 • Published • 27 -
Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning
Paper • 2606.15007 • Published • 19 -
Metis: Memory Foundation Model
Paper • 2607.26760 • Published • 269