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%.
arXiv preprint arXiv:2302.07253 , year=
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 2years
2026 2verdicts
UNVERDICTED 2roles
background 1polarities
background 1representative citing papers
Thesis uses statistical mechanics to study DAM and RBM models for understanding memorization, low-dimensional learning, and adversarial robustness in neural networks.
citing papers explorer
-
Reasoning as Attractor Dynamics: Latent Memory Retrieval via Gibbs-Weighted Energy Minimization
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%.
-
Explaining Machine Learning and Memorization with Statistical Mechanics
Thesis uses statistical mechanics to study DAM and RBM models for understanding memorization, low-dimensional learning, and adversarial robustness in neural networks.