STILL: Selecting Tokens for Intra-Layer Hybrid Attention to Linearize LLMs

arXiv:2602.02180 · cs.LG · Submitted 2026-02-02 · Read on arXiv

cs.LG

Submitted: 2026-02-02

Updated: 2026-10-05

Code: https://github.com/togethercomputer/stripedhyena

Terminology

• Attention Mechanism — 1× in this paper · explained in 3 episodes
• Benchmark — 1× in this paper · explained in 2 episodes
• Consistency — 1× in this paper · explained in 3 episodes
• Distribution Shift — 1× in this paper · explained in 7 episodes
• Large Language Models (LLMs) — 1× in this paper · explained in 6 episodes

Sources

• Llamba: Scaling Distilled Recurrent Models for Efficient Language Processing
• RecurrentGemma: Moving Past Transformers for Efficient Open Language Models
• PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling
• Hybrid Linear Attention Done Right: Efficient Distillation and Effective Architectures for Extremely Long Contexts
• Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
• Griffin: Mixing Gated Linear Recurrences with Local Attention for Efficient Language Models
• Neural Attention Search Linear: Towards Adaptive Token-Level Hybrid Attention Models
• Rectifying Magnitude Neglect in Linear Attention
• Zamba: A Compact 7B SSM Hybrid Model
• RADLADS: Rapid Attention Distillation to Linear Attention Decoders at Scale
• Mamba: Linear-Time Sequence Modeling with Selective State Spaces
• RULER: What's the Real Context Size of Your Long-Context Language Models?
• Mistral 7B
• MiniMax-01: Scaling Foundation Models with Lightning Attention
• Distilling to Hybrid Attention Models via KL-Guided Layer Selection
• Linearizing Large Language Models
• Eagle and Finch: RWKV with Matrix-Valued States and Dynamic Recurrence
• TransNormerLLM: A Faster and Better Large Language Model with Improved TransNormer
• Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
• Retentive Network: A Successor to Transformer for Large Language Models

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks