arXiv subjects

Pietro Miotti

Publications and source records attributed to Pietro Miotti.

4 recordsLinked to original sources

arXiv2026

Diffusion as a Training Curriculum for Timestep-Free Iterative Reasoning

Diffusion models and recursive reasoners are both iterative, but they carry information across iterations differently. We add a persistent hidden state to a diffusion denoiser and remove its timestep conditioning, leaving a single shared update that can be run to arbitrary depth. The result is an anytime solver: accuracy keeps improving with inference depth far beyond the rollout lengths and backpropagation window used in training, reaching 99.90% exact solve on Sudoku-Extreme. We also obtain 98.93% solve rate on Maze-Unique. Surprisingly, progressive denoising is unnecessary at inference: holding corruption at its maximum by replacing every non-clue variable with fresh Gaussian noise at each step retains near-perfect solving and converges to stable solutions. This simple noise-injection mechanism enables a single trajectory to efficiently explore the solution space and settle on the correct answer without parallel rollouts, candidate selection, or external verifiers required by prior reasoning models. Nonetheless, ordered annealed corruption remains critical during training, which suggests that diffusion's primary contribution to our anytime solver is not a sampling procedure at inference, but a denoising training curriculum.

cs.LG↗

arXiv2026

Entropy maximization underlies topology and mechanical properties in dynamic covalent hydrogels

Adding dynamic bonds in polymer networks enables reprocessing and recycling; however the full impact of reversible bonds on dynamic network mechanics remains unclear. We build model dynamic networks and observe substantial deviations from classic theory. We rationalize these findings by considering that bond exchange enables the networks to rearrange and adopt a topology with a higher entropy. This allows us to accurately predict the gel point and elasticity of the dynamic networks. Further, we show by controlling bond exchange that network rearrangement can dramatically alter the mechanical properties, even without loss of bonds.

cond-mat.soft↗

arXiv2026

Mesoscopic Modeling of Dynamic Tetra-PEG Hydrogel Networks

We introduce a mesoscopic model of dynamic Tetra-PEG hydrogel networks based on a hybrid Dissipative Particle Dynamics/Monte Carlo (DPD/MC) approach. Polymer chains are described by Finite Extensible Nonlinear Elastic (FENE) potential, while reversible cross-links are modeled with Morse potential and Monte Carlo bond exchange governed by Bell's force-dependent kinetics. After systematic calibration against theory and experiments, the model reproduces the characteristic Maxwell-like viscoelastic response of these networks. In particular, the relaxation time follows the expected scaling, $\tau_R \propto \tau_b (p - p_{\text{gel}})$, and the simulated storage moduli agree with experimental rheology. The mesoscopic resolution allows for graph-based topological analysis, where Tetra-PEG molecules and cross-links are represented as nodes and edges, providing access to bond distributions, fraction of dangling chains, and size of percolating clusters that are challenging to measure experimentally. Comparison with permanent-network predictions further suggests that dynamic bond exchange can affect bond distributions and delay the formation of a system-spanning cluster. This model bridges macromolecular bond kinetics and macroscopic mechanical properties, providing a complementary tool for rational design of dynamic polymer networks.

cond-mat.soft↗

arXiv2025

Differentiable Logic Cellular Automata: From Game of Life to Pattern Generation

This paper introduces Differentiable Logic Cellular Automata (DiffLogic CA), a novel combination of Neural Cellular Automata (NCA) and Differentiable Logic Gates Networks (DLGNs). The fundamental computation units of the model are differentiable logic gates, combined into a circuit. During training, the model is fully end-to-end differentiable allowing gradient-based training, and at inference time it operates in a fully discrete state space. This enables learning local update rules for cellular automata while preserving their inherent discrete nature. We demonstrate the versatility of our approach through a series of milestones: (1) fully learning the rules of Conway's Game of Life, (2) generating checkerboard patterns that exhibit resilience to noise and damage, (3) growing a lizard shape, and (4) multi-color pattern generation. Our model successfully learns recurrent circuits capable of generating desired target patterns. For simpler patterns, we observe success with both synchronous and asynchronous updates, demonstrating significant generalization capabilities and robustness to perturbations. We make the case that this combination of DLGNs and NCA represents a step toward programmable matter and robust computing systems that combine binary logic, neural network adaptability, and localized processing. This work, to the best of our knowledge, is the first successful application of differentiable logic gate networks in recurrent architectures.

cs.AI↗