Cover Image

Why Neural Network Can Discover Symbolic Structures with Gradient-based Training: An Algebraic and Geometric Foundation for Neurosymbolic Reasoning

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Wang, Peihao, Wang, Zhangyang
Format:	Preprint
Published:	2025
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

$\nabla$-Reasoner: LLM Reasoning via Test-Time Gradient Descent in Latent Space
by: Wang, Peihao, et al.
Published: (2026)

Generalization Error Analysis for Sparse Mixture-of-Experts: A Preliminary Study
by: Zhao, Jinze, et al.
Published: (2024)

Position: Weight Space Should Be a First-Class Generative AI Modality
by: Wang, Zhangyang, et al.
Published: (2026)

On How Iterative Magnitude Pruning Discovers Local Receptive Fields in Fully Connected Neural Networks
by: Redman, William T., et al.
Published: (2024)

Neurosymbolic LoRA: Why and When to Tune Weights vs. Rewrite Prompts
by: Wang, Kevin, et al.
Published: (2026)

When Do Graph Foundation Models Transfer? A Data-Centric Theory
by: Zhu, Jiajun, et al.
Published: (2026)

Graph-KV: Breaking Sequence via Injecting Structural Biases into Large Language Models
by: Wang, Haoyu, et al.
Published: (2025)

Polynomial Width is Sufficient for Set Representation with High-dimensional Features
by: Wang, Peihao, et al.
Published: (2023)

Sparse Mixture-of-Experts for Compositional Generalization: Empirical Evidence and Theoretical Foundations of Optimal Sparsity
by: Zhao, Jinze, et al.
Published: (2024)

You Can Have Better Graph Neural Networks by Not Training Weights at All: Finding Untrained GNNs Tickets
by: Huang, Tianjin, et al.
Published: (2022)

Training Dynamics of Transformers to Recognize Word Co-occurrence via Gradient Flow Analysis
by: Yang, Hongru, et al.
Published: (2024)

Meta ControlNet: Enhancing Task Adaptation via Meta Learning
by: Yang, Junjie, et al.
Published: (2023)

GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection
by: Zhao, Jiawei, et al.
Published: (2024)

Understanding and Mitigating Bottlenecks of State Space Models through the Lens of Recency and Over-smoothing
by: Wang, Peihao, et al.
Published: (2024)

Can Blockchains Reliably Train Machine Learning Models?
by: Li, Peihao, et al.
Published: (2023)

Beyond Test-Time Memory: State-Space Optimal Control for LLM Reasoning
by: Wang, Peihao, et al.
Published: (2026)

Can Neural Networks Learn Small Algebraic Worlds? An Investigation Into the Group-theoretic Structures Learned By Narrow Models Trained To Predict Group Operations
by: Kvinge, Henry, et al.
Published: (2026)

Neurosymbolic Reasoning Shortcuts under the Independence Assumption
by: van Krieken, Emile, et al.
Published: (2025)

Closed-Form Interpretation of Neural Network Classifiers with Symbolic Gradients
by: Wetzel, Sebastian Johann
Published: (2024)

Improving Rule-based Reasoning in LLMs using Neurosymbolic Representations
by: Dhanraj, Varun, et al.
Published: (2025)

Why High-rank Neural Networks Generalize?: An Algebraic Framework with RKHSs
by: Hashimoto, Yuka, et al.
Published: (2025)

BioArc: Discovering Optimal Neural Architectures for Biological Foundation Models
by: Fang, Yi, et al.
Published: (2025)

Discovering Symmetries of ODEs by Symbolic Regression
by: Kahlmeyer, Paul, et al.
Published: (2025)

Symbolic Foundation Regressor on Complex Networks
by: Liu, Weiting, et al.
Published: (2025)

Closed-Form Interpretation of Neural Network Latent Spaces with Symbolic Gradients
by: Wetzel, Sebastian J., et al.
Published: (2024)

Gradient Rewiring for Editable Graph Neural Network Training
by: Jiang, Zhimeng, et al.
Published: (2024)

Discovering Predictive Relational Object Symbols with Symbolic Attentive Layers
by: Ahmetoglu, Alper, et al.
Published: (2023)

Discovering Hidden Algebraic Structures via Transformers with Rank-Aware Beam GRPO
by: Lee, Jaeha, et al.
Published: (2025)

Discovering Symbolic Differential Equations with Symmetry Invariants
by: Yang, Jianke, et al.
Published: (2025)

Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototypical Neurosymbolic AI
by: Andolfi, Luca, et al.
Published: (2025)

Generalized Tangent Kernel: A Unified Geometric Foundation for Natural Gradient and Standard Gradient
by: Bai, Qinxun, et al.
Published: (2022)

Neural Networks with Sparse Activation Induced by Large Bias: Tighter Analysis with Bias-Generalized NTK
by: Yang, Hongru, et al.
Published: (2023)

Beyond Neural Networks: Symbolic Reasoning over Wavelet Logic Graph Signals
by: Kiruluta, Andrew, et al.
Published: (2025)

Gradient-Free Training of Quantized Neural Networks
by: Cohen, Noa, et al.
Published: (2024)

Neural Structure Embedding for Symbolic Regression via Continuous Structure Search and Coefficient Optimization
by: Memar, Fateme, et al.
Published: (2026)

Deep Symbolic Learning: Discovering Symbols and Rules from Perceptions
by: Daniele, Alessandro, et al.
Published: (2022)

The Role of Foundation Models in Neuro-Symbolic Learning and Reasoning
by: Cunnington, Daniel, et al.
Published: (2024)

Imbalances in Neurosymbolic Learning: Characterization and Mitigating Strategies
by: Wang, Kaifu, et al.
Published: (2024)

A Neurosymbolic Framework for Geometric Reduction of Binary Forms
by: Kotsireas, Ilias, et al.
Published: (2025)

Neural Operators Can Discover Functional Clusters
by: Li, Yicen, et al.
Published: (2026)