SketchSSM: Write to the Full State, Read from a Compact Sketch

arXiv:2609.33051 · cs.LG, cs.DC · Submitted 2026-09-27 · Read on arXiv

cs.LG, cs.DC

Submitted: 2026-09-27

Updated: 2026-09-27

Code: https://github.com/Johnny-Liou/ReplaySSM

Terminology

• Decode — 49× in this paper · explained in 2 episodes
• Accuracy — 45× in this paper · explained in 1 episode
• Calibration — 20× in this paper · explained in 7 episodes
• Quantization — 17× in this paper · explained in 12 episodes
• Pruning — 15× in this paper · explained in 4 episodes
• Benchmark — 12× in this paper · explained in 2 episodes
• Ablation — 9× in this paper · explained in 1 episode
• Transformer — 8× in this paper · explained in 2 episodes
• Reconstruction Error — 6× in this paper · explained in 2 episodes
• International Conference on Machine Learning — 5× in this paper
• KV cache — 4× in this paper · explained in 3 episodes
• Sparse Attention — 2× in this paper · explained in 5 episodes
• Sparsity — 2× in this paper · explained in 4 episodes
• Agentic Reasoning — 1× in this paper · explained in 1 episode
• Computational overhead — 1× in this paper · explained in 2 episodes
• Decoupled Scale Quantization — 1× in this paper
• DeepSeek Sparse Attention — 1× in this paper
• Hybrid Models — 1× in this paper · explained in 2 episodes
• Qwen Sparse Attention — 1× in this paper
• Regularization — 1× in this paper · explained in 3 episodes

Sources

• RULER: What's the Real Context Size of Your Long-Context Language Models?
• LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
• Jamba: A Hybrid Transformer-Mamba Language Model
• Let's Verify Step by Step
• The Key to State Reduction in Linear Attention: A Rank-based Perspective
• Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning
• ASVD: Activation-aware Singular Value Decomposition for Compressing Large Language Models
• DAMP: Decay-Aware Mixed-Precision Recurrent-State Quantization
• KVBuffer: IO-aware Serving for Linear Attention

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: Local-Consistency Optimisation for Bayesian Neural Networks
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks