Self-Attention Limits Working Memory Capacity of Transformer-Based Models

• URL: http://arxiv.org/abs/2409.10715v2
• Date: Sat, 16 Nov 2024 20:50:11 GMT
• Title: Self-Attention Limits Working Memory Capacity of Transformer-Based Models
• Authors: Dongyu Gong, Hantao Zhang,
• Abstract summary: Recent work on Transformer-based large language models (LLMs) has revealed striking limits in their working memory capacity. Specifically, these models' performance drops significantly on N-back tasks as N increases. Inspired by the executive attention theory from behavioral sciences, we hypothesize that the self-attention mechanism might be responsible for their working memory capacity limits.
• Score: 0.46040036610482665
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent work on Transformer-based large language models (LLMs) has revealed striking limits in their working memory capacity, similar to what has been found in human behavioral studies. Specifically, these models' performance drops significantly on N-back tasks as N increases. However, there is still a lack of mechanistic interpretability as to why this phenomenon would arise. Inspired by the executive attention theory from behavioral sciences, we hypothesize that the self-attention mechanism within Transformer-based models might be responsible for their working memory capacity limits. To test this hypothesis, we train vanilla decoder-only transformers to perform N-back tasks and find that attention scores gradually aggregate to the N-back positions over training, suggesting that the model masters the task by learning a strategy to pay attention to the relationship between the current position and the N-back position. Critically, we find that the total entropy of the attention score matrix increases as N increases, suggesting that the dispersion of attention scores might be the cause of the capacity limit observed in N-back tasks. Our findings thus offer insights into the shared role of attention in both human and artificial intelligence. Moreover, the limitations of the self-attention mechanism revealed in the current study could inform future efforts to design more powerful model architectures with enhanced working memory capacity and cognitive capabilities.

Related papers

• Langevin Flows for Modeling Neural Latent Dynamics [81.81271685018284]
  We introduce LangevinFlow, a sequential Variational Auto-Encoder where the time evolution of latent variables is governed by the underdamped Langevin equation.<n>Our approach incorporates physical priors -- such as inertia, damping, a learned potential function, and forces -- to represent both autonomous and non-autonomous processes in neural systems.<n>Our method outperforms state-of-the-art baselines on synthetic neural populations generated by a Lorenz attractor.
  arXiv  Detail & Related papers  (2025-07-15T17:57:48Z)
• When Does Neuroevolution Outcompete Reinforcement Learning in Transfer Learning Tasks? [7.479827648985631]
  Neuroevolution (NE) has recently gained attention for its robustness, scalability, and capacity to escape local optima.<n>We introduce two benchmarks: a) in stepping gates, neural networks are tasked with emulating logic circuits, with designs that emphasize modular repetition and variation b) ecorobot extends the Brax physics engine with objects such as walls and obstacles.<n>Our empirical analysis shows that NE methods vary in their transfer abilities and frequently outperform RL baselines.
  arXiv  Detail & Related papers  (2025-05-28T13:15:18Z)
• Language Models Are Capable of Metacognitive Monitoring and Control of Their Internal Activations [1.0485739694839669]
  Large language models (LLMs) can sometimes report the strategies they actually use to solve tasks, but they can also fail to do so.<n>This suggests some degree of metacognition -- the capacity to monitor one's own cognitive processes for subsequent reporting and self-control.<n>We introduce a neuroscience-inspired neurofeedback paradigm designed to quantify the ability of LLMs to explicitly report and control their activation patterns.
  arXiv  Detail & Related papers  (2025-05-19T22:32:25Z)
• Exploring Channel Distinguishability in Local Neighborhoods of the Model Space in Quantum Neural Networks [0.5277756703318045]
  Quantum Neural Networks (QNNs) have emerged and gained significant attention. QNNs have been shown to be notoriously difficult to train, which we hypothesize is partially due to the architectures, called ansatzes.
  arXiv  Detail & Related papers  (2024-10-12T10:20:26Z)
• Modularity in Transformers: Investigating Neuron Separability & Specialization [0.0]
  Transformer models are increasingly prevalent in various applications, yet our understanding of their internal workings remains limited. This paper investigates the modularity and task specialization of neurons within transformer architectures, focusing on both vision (ViT) and language (Mistral 7B) models. Using a combination of selective pruning and MoEfication clustering techniques, we analyze the overlap and specialization of neurons across different tasks and data subsets.
  arXiv  Detail & Related papers  (2024-08-30T14:35:01Z)
• Too Big to Fail: Larger Language Models are Disproportionately Resilient to Induction of Dementia-Related Linguistic Anomalies [7.21603206617401]
  We show that larger GPT-2 models require a disproportionately larger share of attention heads to be masked/ablated to display degradation magnitude to masking. These results suggest that the attention mechanism in transformer models may present an analogue to the notions of cognitive and brain reserve.
  arXiv  Detail & Related papers  (2024-06-05T00:31:50Z)
• Enhancing Generative Class Incremental Learning Performance with Model Forgetting Approach [50.36650300087987]
  This study presents a novel approach to Generative Class Incremental Learning (GCIL) by introducing the forgetting mechanism. We have found that integrating the forgetting mechanisms significantly enhances the models' performance in acquiring new knowledge.
  arXiv  Detail & Related papers  (2024-03-27T05:10:38Z)
• Exploring Model Transferability through the Lens of Potential Energy [78.60851825944212]
  Transfer learning has become crucial in computer vision tasks due to the vast availability of pre-trained deep learning models. Existing methods for measuring the transferability of pre-trained models rely on statistical correlations between encoded static features and task labels. We present an insightful physics-inspired approach named PED to address these challenges.
  arXiv  Detail & Related papers  (2023-08-29T07:15:57Z)
• Understanding Self-attention Mechanism via Dynamical System Perspective [58.024376086269015]
  Self-attention mechanism (SAM) is widely used in various fields of artificial intelligence. We show that intrinsic stiffness phenomenon (SP) in the high-precision solution of ordinary differential equations (ODEs) also widely exists in high-performance neural networks (NN) We show that the SAM is also a stiffness-aware step size adaptor that can enhance the model's representational ability to measure intrinsic SP.
  arXiv  Detail & Related papers  (2023-08-19T08:17:41Z)
• Learning Theory of Mind via Dynamic Traits Attribution [59.9781556714202]
  We propose a new neural ToM architecture that learns to generate a latent trait vector of an actor from the past trajectories. This trait vector then multiplicatively modulates the prediction mechanism via a fast weights' scheme in the prediction neural network. We empirically show that the fast weights provide a good inductive bias to model the character traits of agents and hence improves mindreading ability.
  arXiv  Detail & Related papers  (2022-04-17T11:21:18Z)
• Reducing Catastrophic Forgetting in Self Organizing Maps with Internally-Induced Generative Replay [67.50637511633212]
  A lifelong learning agent is able to continually learn from potentially infinite streams of pattern sensory data. One major historic difficulty in building agents that adapt is that neural systems struggle to retain previously-acquired knowledge when learning from new samples. This problem is known as catastrophic forgetting (interference) and remains an unsolved problem in the domain of machine learning to this day.
  arXiv  Detail & Related papers  (2021-12-09T07:11:14Z)
• Attention or memory? Neurointerpretable agents in space and time [0.0]
  We design a model incorporating a self-attention mechanism that implements task-state representations in semantic feature-space. To evaluate the agent's selective properties, we add a large volume of task-irrelevant features to observations. In line with neuroscience predictions, self-attention leads to increased robustness to noise compared to benchmark models.
  arXiv  Detail & Related papers  (2020-07-09T15:04:26Z)
• Untangling tradeoffs between recurrence and self-attention in neural networks [81.30894993852813]
  We present a formal analysis of how self-attention affects gradient propagation in recurrent networks. We prove that it mitigates the problem of vanishing gradients when trying to capture long-term dependencies. We propose a relevancy screening mechanism that allows for a scalable use of sparse self-attention with recurrence.
  arXiv  Detail & Related papers  (2020-06-16T19:24:25Z)

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.