English
Related papers

Related papers: Stochastic Thermodynamics for Autoregressive Gener…

200 papers

Traditional supervised fine-tuning (SFT) strategies for sequence-to-sequence tasks often train models to directly generate the target output. Recent work has shown that guiding models with intermediate steps, such as keywords, outlines, or…

Computation and Language · Computer Science 2025-02-19 Senyu Li , Zipeng Sun , Jiayi Wang , Xue Liu , Pontus Stenetorp , Siva Reddy , David Ifeoluwa Adelani

We derive an Ito stochastic differential equation for entropy production in nonequilibrium Langevin processes. Introducing a random-time transformation, entropy production obeys a one-dimensional drift-diffusion equation, independent of the…

Statistical Mechanics · Physics 2017-10-10 Simone Pigolotti , Izaak Neri , Édgar Roldán , Frank Jülicher

We study dynamical reversibility in stationary stochastic processes from an information theoretic perspective. Extending earlier work on the reversibility of Markov chains, we focus on finitary processes with arbitrarily long conditional…

Statistical Mechanics · Physics 2015-05-28 Christopher J. Ellison , John R. Mahoney , Ryan G. James , James P. Crutchfield , Joerg Reichardt

Certain thermal non-equilibrium situations, outside of the astrophysical realm, suggest that entropy production extrema, instead of entropy extrema, are related to stationary states. In an effort to better understand the evolution of…

Astrophysics of Galaxies · Physics 2015-05-27 Eric I. Barnes , Liliya L. R. Williams

We consider a stochastic process which is (a) described by a continuous-time Markov chain on only short time-scales and (b) constrained to conserve a number of hidden quantities on long time-scales. We assume that the transition matrix of…

Statistical Mechanics · Physics 2020-10-27 Vitaly Vanchurin

Spatial-Temporal Graph (STG) data is characterized as dynamic, heterogenous, and non-stationary, leading to the continuous challenge of spatial-temporal graph learning. In the past few years, various GNN-based methods have been proposed to…

Machine Learning · Computer Science 2024-05-21 Lincan Li , Hanchen Wang , Wenjie Zhang , Adelle Coster

Non-Markovian models have great expressive power, at the cost of complex analysis of the stochastic process. The method of Stochastic State Classes (SSCs) derives closed-form analytical expressions for the joint Probability Density…

Logic in Computer Science · Computer Science 2024-10-24 Gabriel Dengler , Laura Carnevali , Carlos E. Budde , Enrico Vicario

Entropy production (EP) is a central measure in nonequilibrium thermodynamics, as it can quantify the irreversibility of a process as well as its energy dissipation in special cases. Using the time-reversal asymmetry in a system's path…

Statistical Mechanics · Physics 2022-04-26 Dong-Kyum Kim , Sangyun Lee , Hawoong Jeong

Diffusion models (DMs) represent state-of-the-art generative models for continuous inputs. DMs work by constructing a Stochastic Differential Equation (SDE) in the input space (ie, position space), and using a neural network to reverse it.…

Machine Learning · Computer Science 2024-05-14 Tianrong Chen , Jiatao Gu , Laurent Dinh , Evangelos A. Theodorou , Joshua Susskind , Shuangfei Zhai

We incorporate discrete and continuous time Markov processes as building blocks into probabilistic graphical models with latent and observed variables. We introduce the automatic Backward Filtering Forward Guiding (BFFG) paradigm (Mider et…

Computation · Statistics 2022-11-02 Frank van der Meulen , Moritz Schauer

Diffusion language models (DLMs) promise parallel, order-agnostic generation, but on standard benchmarks they have historically lagged behind autoregressive models in sample quality and diversity. Recent continuous flow and diffusion…

Computation and Language · Computer Science 2026-05-11 Georgios Batzolis , Mark Girolami , Luca Ambrogioni

We focus on generative autoencoders, such as variational or adversarial autoencoders, which jointly learn a generative model alongside an inference model. Generative autoencoders are those which are trained to softly enforce a prior on the…

Machine Learning · Computer Science 2017-01-13 Antonia Creswell , Kai Arulkumaran , Anil Anthony Bharath

Traditional thermodynamic trade-off relations usually apply to quantities that depend linearly on probability distributions. In contrast, many important information-theoretic measures, such as entropies, are nonlinear and therefore…

Statistical Mechanics · Physics 2026-02-17 Yoshihiko Hasegawa

We present an analysis of two thermodynamic techniques for determining equilibria of self-gravitating systems. One is the Lynden-Bell entropy maximization analysis that introduced violent relaxation. Since we do not use the Stirling…

Cosmology and Nongalactic Astrophysics · Physics 2015-06-03 Eric I. Barnes , Liliya L. R. Williams

We develop a new thermodynamic approach to stochastic graph-rewriting. The ingredients are a finite set of reversible graph-rewriting rules called generating rules, a finite set of connected graphs P called energy patterns and an energy…

Logic in Computer Science · Computer Science 2017-01-11 Vincent Danos , Russell Harmer , Ricardo Honorato-Zimmer

We introduce a novel class of generative models based on piecewise deterministic Markov processes (PDMPs), a family of non-diffusive stochastic processes consisting of deterministic motion and random jumps at random times. Similarly to…

Machine Learning · Statistics 2024-11-06 Andrea Bertazzi , Dario Shariatian , Umut Simsekli , Eric Moulines , Alain Durmus

Modeling the time-dependent evolution of electron density is essential for understanding quantum mechanical behaviors of condensed matter and enabling predictive simulations in spectroscopy, photochemistry, and ultrafast science. Yet, while…

Computational Physics · Physics 2025-09-03 Yuan Chiang , Youngsoo Choi , Daniel Osei-Kuffuor

Based on some recent work of the author on stochastic approximation in non-markovian environments, the situation when the driving random process is non-ergodic in addition to being non-markovian is considered. Using this, we propose an…

Machine Learning · Statistics 2026-03-24 Vivek Shripad Borkar

We consider optimization of the average entropy production in inhomogeneous temperature environments within the framework of stochastic thermodynamics. For systems modeled by Langevin equations (e.g. a colloidal particle in a heat bath) it…

Statistical Mechanics · Physics 2013-03-14 Stefano Bo , Erik Aurell , Ralf Eichhorn , Antonio Celani

Generative transformer models have become increasingly complex, with large numbers of parameters and the ability to process multiple input modalities. Current methods for explaining their predictions are resource-intensive. Most crucially,…

Machine Learning · Computer Science 2025-01-08 Björn Deiseroth , Mayukh Deb , Samuel Weinbach , Manuel Brack , Patrick Schramowski , Kristian Kersting