L$^3$: Large Lookup Layers

• URL: http://arxiv.org/abs/2601.21461v2
• Date: Fri, 30 Jan 2026 09:13:05 GMT
• Title: L$^3$: Large Lookup Layers
• Authors: Albert Tseng, Christopher De Sa,
• Abstract summary: A sparse tokenizer embedding table selects a single embedding per token at the cost of not having contextual information.<n>L$3$ layers use static token-based routing to aggregate a set of learned embeddings per token in a context-dependent way.<n>L$3$ has two main components: (1) a systems-friendly architecture that allows for fast training and CPU-offloaded inference with no overhead, and (2) an information-theoretic embedding allocation algorithm that effectively balances speed and quality.
• Score: 30.519997678052658
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Modern sparse language models typically achieve sparsity through Mixture-of-Experts (MoE) layers, which dynamically route tokens to dense MLP "experts." However, dynamic hard routing has a number of drawbacks, such as potentially poor hardware efficiency and needing auxiliary losses for stable training. In contrast, the tokenizer embedding table, which is natively sparse, largely avoids these issues by selecting a single embedding per token at the cost of not having contextual information. In this work, we introduce the Large Lookup Layer (L$^3$), which unlocks a new axis of sparsity by generalizing embedding tables to model decoder layers. L$^3$ layers use static token-based routing to aggregate a set of learned embeddings per token in a context-dependent way, allowing the model to efficiently balance memory and compute by caching information in embeddings. L$^3$ has two main components: (1) a systems-friendly architecture that allows for fast training and CPU-offloaded inference with no overhead, and (2) an information-theoretic embedding allocation algorithm that effectively balances speed and quality. We empirically test L$^3$ by training transformers with up to 2.6B active parameters and find that L$^3$ strongly outperforms both dense models and iso-sparse MoEs in both language modeling and downstream tasks.

Related papers

• Stacked from One: Multi-Scale Self-Injection for Context Window Extension [69.24689919827817]
  modelname is a novel framework based on multi-grained context compression and query-aware information acquisition.<n>modelnameachieves performance superior or comparable to strong baselines.
  arXiv  Detail & Related papers  (2026-03-05T03:16:16Z)
• SimpleMem: Efficient Lifelong Memory for LLM Agents [73.74399447715052]
  We introduce SimpleMem, an efficient memory framework based on semantic lossless compression.<n>We propose a three-stage pipeline designed to maximize information density and token utilization.<n> Experiments on benchmark datasets show that our method consistently outperforms baseline approaches in accuracy, retrieval efficiency, and inference cost.
  arXiv  Detail & Related papers  (2026-01-05T21:02:49Z)
• Mixture of States: Routing Token-Level Dynamics for Multimodal Generation [60.15447534872979]
  We introduce MoS (Mixture of States), a novel fusion paradigm for multimodal diffusion models.<n>MoS is a learnable, token-wise router that creates denoising timestep- and input-dependent interactions.<n>We validate our design with text-to-image generation and editing, which achieve state-of-the-art results.
  arXiv  Detail & Related papers  (2025-11-15T13:24:57Z)
• DeepInsert: Early Layer Bypass for Efficient and Performant Multimodal Understanding [26.39397960987363]
  We propose a simple modification to pretrained transformer models.<n>Instead of concatenation with the language prompt at the start, we insert multimodal tokens directly into the middle.<n>Our results indicate that our method reduces computational costs during both training and inference.
  arXiv  Detail & Related papers  (2025-04-27T18:56:26Z)
• Saliency-driven Dynamic Token Pruning for Large Language Models [32.903622070917194]
  Saliency-driven Dynamic Token Pruning (SDTP)<n>A lightweight saliency-driven prediction module is designed to estimate the importance score of each token with its hidden state.<n>A ranking-based optimization strategy is proposed to minimize the ranking divergence of the saliency score and the predicted importance score.
  arXiv  Detail & Related papers  (2025-04-06T15:15:07Z)
• RSQ: Learning from Important Tokens Leads to Better Quantized LLMs [65.5558181902098]
  Layer-wise quantization is a key technique for efficiently compressing large models without expensive retraining.<n>We propose RSQ (Rotate, Scale, then Quantize), which applies rotations to the model to mitigate outliers.<n>We demonstrate that RSQ consistently outperforms baseline methods across multiple downstream tasks and three model families.
  arXiv  Detail & Related papers  (2025-03-03T18:46:33Z)
• Scaling Embedding Layers in Language Models [61.939921364422936]
  $SCONE$ is a new method for extending input embedding layers to enhance language model performance.<n>$SCONE$ retains the original vocabulary while introducing embeddings for a set of frequent n-grams.<n>These embeddings provide contextualized representation for each input token and are learned with a separate model during training.<n>$SCONE$ enables two new scaling strategies: increasing the number of n-gram embeddings and scaling the model used to learn them, both while maintaining fixed accelerator usage during inference.
  arXiv  Detail & Related papers  (2025-02-03T18:59:32Z)
• ST$^3$: Accelerating Multimodal Large Language Model by Spatial-Temporal Visual Token Trimming [14.937905258757635]
  $textbfST3$ is a framework designed to accelerate MLLM inference without retraining.<n>$textbfST3$ can be seamlessly integrated into existing pre-trained MLLMs.
  arXiv  Detail & Related papers  (2024-12-28T10:17:29Z)
• Dynamic layer selection in decoder-only transformers [21.18795712840146]
  We empirically examine two common dynamic inference methods for natural language generation. We find that a pre-trained decoder-only model is significantly more robust to layer removal via layer skipping. We also show that dynamic computation allocation on a per-sequence basis holds promise for significant efficiency gains.
  arXiv  Detail & Related papers  (2024-10-26T00:44:11Z)
• Unveiling Induction Heads: Provable Training Dynamics and Feature Learning in Transformers [54.20763128054692]
  We study how a two-attention-layer transformer is trained to perform ICL on $n$-gram Markov chain data. We prove that the gradient flow with respect to a cross-entropy ICL loss converges to a limiting model.
  arXiv  Detail & Related papers  (2024-09-09T18:10:26Z)
• (PASS) Visual Prompt Locates Good Structure Sparsity through a Recurrent HyperNetwork [60.889175951038496]
  Large-scale neural networks have demonstrated remarkable performance in different domains like vision and language processing. One of the key questions of structural pruning is how to estimate the channel significance. We propose a novel algorithmic framework, namely textttPASS. It is a tailored hyper-network to take both visual prompts and network weight statistics as input, and output layer-wise channel sparsity in a recurrent manner.
  arXiv  Detail & Related papers  (2024-07-24T16:47:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.