FactorEngram: Factorized N-gram Memory with Basis-Level Gating for Language Models

arXiv:2609.35578 · cs.CL, cs.AI · Submitted 2026-09-28 · Read on arXiv

cs.CL, cs.AI

Submitted: 2026-09-28

Updated: 2026-09-28

Code: https://github.com/google-deepmind/gemma

Terminology

• Sparsity — 26× in this paper · explained in 4 episodes
• Transformer — 25× in this paper · explained in 2 episodes
• Accuracy — 24× in this paper · explained in 1 episode
• Embeddings — 15× in this paper · explained in 2 episodes
• Perplexity — 12× in this paper · explained in 3 episodes
• Regularization — 11× in this paper · explained in 3 episodes
• Ablation — 8× in this paper · explained in 1 episode
• Aggregation — 6× in this paper · explained in 1 episode
• Ablation Studies — 3× in this paper · explained in 1 episode
• Per-Layer Embeddings — 2× in this paper
• Adam — 1× in this paper · explained in 2 episodes
• Aggregation Function — 1× in this paper · explained in 1 episode
• Interpretability — 1× in this paper · explained in 15 episodes
• KV cache — 1× in this paper · explained in 3 episodes

Sources

• Memory Grafting: Scaling Language Model Pre-training via Offline Conditional Memory
• Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models
• Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
• A Collision-Free Hot-Tier Extension for Engram-Style Conditional Memory: A Controlled Study of Training Dynamics
• Scaling Embeddings Outperforms Scaling Experts in Language Models
• TF-Engram: A Train-Free Engram with SSD-Backed Memory for Large Language Models
• Pointer Sentinel Mixture Models
• STEM: Scaling Transformers with Embedding Modules
• GLU Variants Improve Transformer
• DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression
• Lngram: N-gram Conditional Memory in Latent Space

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering