Factual recall in linear associative memories: sharp asymptotics and mechanistic insights

• URL: http://arxiv.org/abs/2605.10795v1
• Date: Mon, 11 May 2026 16:24:02 GMT
• Title: Factual recall in linear associative memories: sharp asymptotics and mechanistic insights
• Abstract summary: We study the limits of storing and retrieving input--output associations with neural networks.<n>A linear associative memory maps $p$ input embeddings in $mathbbRd$ to their corresponding$d$-dimensional targets.<n>We show that a decoupled model can store up to $p_c log p_c / d2 = 1 / 2$ associations.
• Score: 11.11958822145404
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models demonstrate remarkable ability in factual recall, yet the fundamental limits of storing and retrieving input--output associations with neural networks remain unclear. We study these limits in a minimal setting: a linear associative memory that maps $p$ input embeddings in $\mathbb{R}^d$ to their corresponding~$d$-dimensional targets via a single layer, requiring each mapped input to be well separated from all other targets. Unlike in supervised classification, this strict separation induces~$p$ constraints per association and produces strong correlations between constraints that make a direct characterisation of the storage capacity difficult. Here, we provide a precise characterisation of this capacity in the following way. We first introduce a decoupled model in which each input has its own independent set of competing outputs, and provide numerical and analytical evidence that this decoupled model is equivalent to the original model in terms of storage capacity, spectra of the learnt weights, and storage mechanism. Using tools from statistical physics, we show that the decoupled model can store up to $p_c \log p_c / d^2 = 1 / 2$ associations, and generalise the computation of $p_c$ to linear two-layer architectures. Our analysis also gives mechanistic insight into how the optimal solution improves over a naïve Hebbian learning rule: rather than boosting input-output alignments with broad fluctuations, the optimal solution raises the correct scores just above the extreme-value threshold set by the competing outputs. These findings give a sharp statistical-physics characterisation of factual storage in linear networks and provide a baseline for understanding the memory capacity of more realistic neural architectures.

Related papers

• Geometric Factual Recall in Transformers [57.48371649045765]
  A common view casts internal weight matrices as associative memories over pairs of embeddings, requiring memorization counts that scale linearly with the number of facts.<n>We develop a theoretical and empirical account of an alternative, emphgeometric form of memorization in which learned embeddings encode structure directly.<n>In a controlled setting where a single-layer transformer must memorize random vectors from subjects to a shared attribute set, we prove that a logarithmic embedding dimension suffices.<n>We extend these results to the multi-hop setting, providing constructions with and without chain-of-thought that exhibit a provable
  arXiv  Detail & Related papers  (2026-05-12T17:22:22Z)
• CAWI: Copula-Aligned Weight Initialization for Randomized Neural Networks [7.698602367824994]
  We propose CAWI (Copula-Aligned Weight Initialization), a framework that draws input-to-hidden weights from a data-fitted copula.<n>We evaluate CAWI across 83 diverse classification benchmarks and two biomedical datasets.
  arXiv  Detail & Related papers  (2026-05-12T15:46:23Z)
• Adaptive Memory Decay for Log-Linear Attention [1.0099625992507715]
  Sequence models face a fundamental tradeoff between memory capacity and computational efficiency.<n>Log-linear attention navigates this tradeoff by organizing memory across a Fenwick tree hierarchy.<n>We propose learning directly from the input via a lightweight two-layer, producing per-token, per-level decay that adapts to content rather than position.
  arXiv  Detail & Related papers  (2026-05-07T21:05:28Z)
• LINA: Linear Autoregressive Image Generative Models with Continuous Tokens [56.80443965097921]
  Autoregressive models with continuous tokens form a promising paradigm for visual generation, especially for text-to-image (T2I) synthesis.<n>We study how to design compute-efficient linear attention within this framework.<n>We present LINA, a simple and compute-efficient T2I model built entirely on linear attention, capable of generating high-fidelity 1024x1024 images from user instructions.
  arXiv  Detail & Related papers  (2026-01-30T06:44:33Z)
• Boundary-Guided Policy Optimization for Memory-efficient RL of Diffusion Large Language Models [53.339700196282905]
  A key challenge in applying reinforcement learning to large language models (dLLMs) is the intractability of their likelihood functions.<n>We propose a memory-efficient RL algorithm that maximizes a specially constructed lower bound of the ELBO-based objective.<n> Experiments show that BGPO significantly outperforms previous RL algorithms for dLLMs in math problem solving, code generation, and planning tasks.
  arXiv  Detail & Related papers  (2025-10-13T17:47:50Z)
• Bound by semanticity: universal laws governing the generalization-identification tradeoff [8.437463955457423]
  We show that finite-resolution similarity is a fundamental emergent informational constraint, not merely a toy-model artifact.<n>These results provide an exact theory of the generalization-identification trade-off and clarify how semantic resolution shapes the representational capacity of deep networks and brains alike.
  arXiv  Detail & Related papers  (2025-06-01T15:56:26Z)
• Bilinear Convolution Decomposition for Causal RL Interpretability [0.0]
  Efforts to interpret reinforcement learning (RL) models often rely on high-level techniques such as attribution or probing.<n>This work proposes replacing nonlinearities in convolutional neural networks (ConvNets) with bilinear variants, to produce a class of models for which these limitations can be addressed.<n>We show bilinear model variants perform comparably in model-free reinforcement learning settings, and give a side by side comparison on ProcGen environments.
  arXiv  Detail & Related papers  (2024-12-01T19:32:04Z)
• 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)
• Implicit Bias and Fast Convergence Rates for Self-attention [26.766649949420746]
  We study the fundamental optimization principles of self-attention, the defining mechanism of transformers.<n>We analyze the implicit bias of gradient-baseds in a self-attention layer with a decoder in a linear classification.
  arXiv  Detail & Related papers  (2024-02-08T15:15:09Z)
• Disentanglement via Latent Quantization [60.37109712033694]
  In this work, we construct an inductive bias towards encoding to and decoding from an organized latent space. We demonstrate the broad applicability of this approach by adding it to both basic data-re (vanilla autoencoder) and latent-reconstructing (InfoGAN) generative models.
  arXiv  Detail & Related papers  (2023-05-28T06:30:29Z)
• FeDXL: Provable Federated Learning for Deep X-Risk Optimization [105.17383135458897]
  We tackle a novel federated learning (FL) problem for optimizing a family of X-risks, to which no existing algorithms are applicable. The challenges for designing an FL algorithm for X-risks lie in the non-decomability of the objective over multiple machines and the interdependency between different machines.
  arXiv  Detail & Related papers  (2022-10-26T00:23:36Z)
• Relative gradient optimization of the Jacobian term in unsupervised deep learning [9.385902422987677]
  Learning expressive probabilistic models correctly describing the data is a ubiquitous problem in machine learning. Deep density models have been widely used for this task, but their maximum likelihood based training requires estimating the log-determinant of the Jacobian. We propose a new approach for exact training of such neural networks.
  arXiv  Detail & Related papers  (2020-06-26T16:41:08Z)

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.