STEM: SCALING TRANSFORMERS WITH EMBEDDING MODULES

ICLR 2026 Conference Submission21909 Authors

Keywords: Sparse Transformer, Parametric scaling, Embedding Layers, Foundation Models, Pre-training, Model Architecture

TL;DR: STEM replaces each FFN up-projection with a per-layer embedding lookup to scale parametric capacity without increasing per-token compute or cross-device communication, yielding FLOP-efficient performance gains.

Abstract: Fine-grained sparsity promises higher parametric capacity without proportional per-token compute, but often suffers from training instability, load balancing, and communication overhead. We introduce \textbf{STEM} (\emph{Scaling Transformers with Embedding Modules}), a static, token-indexed approach that replaces the FFN up-projection with a layer-local embedding lookup while keeping the gate and down-projection dense. This removes runtime routing, enables CPU offload with asynchronous prefetch, and decouples capacity from both per-token FLOPs and cross-device communication. Empirically, STEM trains stably despite extreme sparsity. It improves downstream performance over dense baselines while reducing per-token FLOPs and parameter accesses (eliminating roughly one-third of FFN parameters). STEM learns embedding spaces with large angular spread which enhances it knowledge storage capacity. In addition, STEM strengthens long-context performance: as sequence length grows, more distinct parameters are activated, yielding practical test-time capacity scaling. Across 350M and 1B model scales, STEM delivers up to $\sim$3--4\% accuracy improvements overall, with notable gains on knowledge and reasoning-heavy benchmarks (ARC-Challenge, OpenBookQA, GSM8K, MMLU). Overall, STEM is an effective way of scaling parametric memory while remaining simpler to train and deploy than existing fine-grained sparse models.

Primary Area: foundation or frontier models, including LLMs

Submission Number: 21909