neural-symbolic

NeST-Bayes是用于多机器人系统的神经-符号信任推断框架，仅凭跨机器人观测一致性联合推断机器人与物体假设的信任度，无需预设信任关系，并在对抗攻击与遮挡环境下取得高精度和召回率。

该论文探讨了将经典符号式文档自动化（如专家系统风格的法律文档工具）与生成式AI相结合的神经符号方法，并报告了使用大语言模型识别和修复普通人撰写的法律文本中问题的初步实验。

The paper introduces Coffee, a plug-and-play framework that guides discrete diffusion models at inference time using compiled finite-state objectives, avoiding exponential enumeration of token completions while supporting hard constraints and learned soft objectives across symbolic, language, and biological benchmarks.

This paper introduces MT-PDCL, a measure-theoretic probabilistic definite clause logic framework that generalizes probabilistic logic programming to continuous domains by using Lebesgue integration over standard Borel σ-algebras instead of discrete grounding. It replaces combinatorial grounding bottlenecks with exact algebraic and differentiable inference while preserving declarative definite clause syntax.

This paper presents a three-stage neural-symbolic pipeline for gaming toxicity detection, combining transformer ensembles with rule-based mediation, achieving top accuracy in the EEUCA 2026 shared task.

This paper introduces NeurOWL, a neuro-symbolic framework using LLMs and ontology embeddings to perform subsumption verification and abduction on incomplete OWL ontologies, enabling reasoning when axioms are missing.

This paper proposes a method to extract an executable Prolog program from a deep reinforcement learning policy, providing theoretical guarantees on return and fidelity, enabling interpretability and manual editing.

Presents FormalAnalyticGeo, a neural-symbolic framework for automatic generation of multimodal analytic geometry problems using a formal intermediate representation (CDL) and LLM components, yielding the AnalyticGeo7K dataset of over 7K verified problems with high geometric precision.

Current mainstream pure data-driven robot solutions suffer from low data efficiency and poor generalization. The newly proposed neuro-symbolic physical intelligence paradigm breaks down tasks into two steps: world modeling and planning. It requires only 1-10 demonstrations to learn new tasks, and its generalization ability far exceeds traditional end-to-end solutions, providing a more reliable path for general-purpose robots.

This paper proposes a new security paradigm for AI agents using a Proof-Constrained Action (ePCA) framework with neural symbolic isolation, achieving zero attack success rate in empirical evaluations.

The paper introduces Chimera Training, a method for logical anomaly detection that uses counterfactual construction at the feature level to train neural rule evaluators without requiring real anomalous images, improving rule-level anomaly detection performance on benchmarks like CLEVRER, OpenImages, and VidOR.