Pith. sign in

REVIEW 3 cited by

Energy Transformer

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2302.07253 v2 pith:P2XAB4YY submitted 2023-02-14 cs.LG cond-mat.dis-nncs.CVq-bio.NCstat.ML

Energy Transformer

classification cs.LG cond-mat.dis-nncs.CVq-bio.NCstat.ML
keywords energyattentionmodelstheoreticalallowassociativedesignenergy-based
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Our work combines aspects of three promising paradigms in machine learning, namely, attention mechanism, energy-based models, and associative memory. Attention is the power-house driving modern deep learning successes, but it lacks clear theoretical foundations. Energy-based models allow a principled approach to discriminative and generative tasks, but the design of the energy functional is not straightforward. At the same time, Dense Associative Memory models or Modern Hopfield Networks have a well-established theoretical foundation, and allow an intuitive design of the energy function. We propose a novel architecture, called the Energy Transformer (or ET for short), that uses a sequence of attention layers that are purposely designed to minimize a specifically engineered energy function, which is responsible for representing the relationships between the tokens. In this work, we introduce the theoretical foundations of ET, explore its empirical capabilities using the image completion task, and obtain strong quantitative results on the graph anomaly detection and graph classification tasks.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Competition, stability, and functionality in excitatory-inhibitory neural circuits

    q-bio.NC 2025-12 conditional novelty 5.0

    Asymmetric E-I firing-rate dynamics are reframed as a multi-neuron game, and lateral-inhibition columns are shown to sharpen subthreshold contrasts with a closed-form depth requirement.

  2. Reasoning as Attractor Dynamics: Latent Memory Retrieval via Gibbs-Weighted Energy Minimization

    cs.LG 2026-06 unverdicted novelty 4.0

    LLMs are modeled as dense associative memories with reasoning as attractor dynamics; Gibbs-weighted sampling by spectral entropy improves Phi-3.5 on GSM8K from 84.7% to 90.1%.

  3. Explaining Machine Learning and Memorization with Statistical Mechanics

    cs.LG 2026-06 unverdicted novelty 3.0

    Thesis uses statistical mechanics to study DAM and RBM models for understanding memorization, low-dimensional learning, and adversarial robustness in neural networks.