Blending Complementary Memory Systems in Hybrid Quadratic-Linear Transformers

• URL: http://arxiv.org/abs/2506.00744v1
• Date: Sat, 31 May 2025 23:16:53 GMT
• Title: Blending Complementary Memory Systems in Hybrid Quadratic-Linear Transformers
• Authors: Kazuki Irie, Morris Yau, Samuel J. Gershman,
• Abstract summary: We develop hybrid memory architectures for general-purpose sequence processing neural networks.<n>We combine key-value memory using softmax attention ( KV-memory) with dynamic synaptic memory through fast-weight programming (FW-memory)<n>We show how a well-designed hybrid can overcome the limitations of its individual components.
• Score: 14.130531534945577
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We develop hybrid memory architectures for general-purpose sequence processing neural networks, that combine key-value memory using softmax attention (KV-memory) with dynamic synaptic memory through fast-weight programming (FW-memory) -- the core principles of quadratic and linear transformers, respectively. These two memory systems have complementary but individually limited properties: KV-memory offers precise retrieval but is constrained by quadratic complexity in sequence length, while FW-memory supports arbitrarily long sequences and enables more expressive computation but sacrifices precise recall. We propose and compare three methods to blend these two systems into a single memory system to leverage the strengths of both. We conduct experiments on general language modeling and retrieval tasks by training 340M- and 1.3B-parameter models from scratch, as well as on synthetic algorithmic tasks designed to precisely illustrate the benefits of certain hybrid methods over others. We also evaluate our hybrid memory systems on reinforcement learning in partially observable environments. Overall, we demonstrate how a well-designed hybrid can overcome the limitations of its individual components, offering new insights into the design principle of neural memory systems.

Related papers

• mGRADE: Minimal Recurrent Gating Meets Delay Convolutions for Lightweight Sequence Modeling [0.5236468296934584]
  mGRADE is a hybrid-memory system that integrates a temporal 1D-convolution with learnable spacings followed by a minimal gated recurrent unit.<n>We demonstrate that mGRADE effectively separates and preserves multi-scale temporal features.<n>This highlights mGRADE's promise as an efficient solution for memory-constrained multi-scale temporal processing at the edge.
  arXiv  Detail & Related papers  (2025-07-02T15:44:35Z)
• Hardware-software co-exploration with racetrack memory based in-memory computing for CNN inference in embedded systems [54.045712360156024]
  racetrack memory is a non-volatile technology that allows high data density fabrication.<n>In-memory arithmetic circuits with memory cells affects both the memory density and power efficiency.<n>We present an efficient in-memory convolutional neural network (CNN) accelerator optimized for use with racetrack memory.
  arXiv  Detail & Related papers  (2025-07-02T07:29:53Z)
• Memory Layers at Scale [67.00854080570979]
  This work takes memory layers beyond proof-of-concept, proving their utility at contemporary scale.<n>On downstream tasks, language models augmented with our improved memory layer outperform dense models with more than twice the budget, as well as mixture-of-expert models when matched for both compute and parameters.<n>We provide a fully parallelizable memory layer implementation, demonstrating scaling laws with up to 128B memory parameters, pretrained to 1 trillion tokens, comparing to base models with up to 8B parameters.
  arXiv  Detail & Related papers  (2024-12-12T23:56:57Z)
• B'MOJO: Hybrid State Space Realizations of Foundation Models with Eidetic and Fading Memory [91.81390121042192]
  We develop a class of models called B'MOJO to seamlessly combine eidetic and fading memory within an composable module. B'MOJO's ability to modulate eidetic and fading memory results in better inference on longer sequences tested up to 32K tokens.
  arXiv  Detail & Related papers  (2024-07-08T18:41:01Z)
• Heterogenous Memory Augmented Neural Networks [84.29338268789684]
  We introduce a novel heterogeneous memory augmentation approach for neural networks. By introducing learnable memory tokens with attention mechanism, we can effectively boost performance without huge computational overhead. We show our approach on various image and graph-based tasks under both in-distribution (ID) and out-of-distribution (OOD) conditions.
  arXiv  Detail & Related papers  (2023-10-17T01:05:28Z)
• Universal Hopfield Networks: A General Framework for Single-Shot Associative Memory Models [41.58529335439799]
  We propose a general framework for understanding the operation of memory networks as a sequence of three operations. We derive all these memory models as instances of our general framework with differing similarity and separation functions.
  arXiv  Detail & Related papers  (2022-02-09T16:48:06Z)
• Semantically Constrained Memory Allocation (SCMA) for Embedding in Efficient Recommendation Systems [27.419109620575313]
  A key challenge for deep learning models is to work with millions of categorical classes or tokens. We propose a novel formulation of memory shared embedding, where memory is shared in proportion to the overlap in semantic information. We demonstrate a significant reduction in the memory footprint while maintaining performance.
  arXiv  Detail & Related papers  (2021-02-24T19:55:49Z)
• Memformer: A Memory-Augmented Transformer for Sequence Modeling [55.780849185884996]
  We present Memformer, an efficient neural network for sequence modeling. Our model achieves linear time complexity and constant memory space complexity when processing long sequences.
  arXiv  Detail & Related papers  (2020-10-14T09:03:36Z)
• Robust High-dimensional Memory-augmented Neural Networks [13.82206983716435]
  Memory-augmented neural networks enhance neural networks with an explicit memory to overcome these issues. Access to this explicit memory occurs via soft read and write operations involving every individual memory entry. We propose a robust architecture that employs a computational memory unit as the explicit memory performing analog in-memory computation on high-dimensional (HD) vectors.
  arXiv  Detail & Related papers  (2020-10-05T12:01:56Z)
• Self-Attentive Associative Memory [69.40038844695917]
  We propose to separate the storage of individual experiences (item memory) and their occurring relationships (relational memory) We achieve competitive results with our proposed two-memory model in a diversity of machine learning tasks.
  arXiv  Detail & Related papers  (2020-02-10T03:27:48Z)

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.