Latest Neuro-Symbolic AI Research Papers
The newest Neuro-Symbolic AI papers from across the field — arXiv, NeurIPS, CVPR, Nature, and more — refreshed daily and ranked by relevance. Distill AI tracks Neuro-Symbolic AI so you don’t have to: get the standout work delivered to your inbox every morning, with 2-sentence summaries and the option to chat with any paper.
Get the latest Neuro-Symbolic AI papers in your inbox — free →Recent papers
- SoftReason: A Fully Differentiable Neuro-Soft-Symbolic Deductive Reasoning Architecture over High-Dimensional Perceptual DataWael AbdAlmageed · arXiv · Jul 22, 2026
In many reasoning problems, the premises are not observed as discrete symbols, but must be inferred from high-dimensional inputs. Further, the predicate vocabulary, argument structure, and trusted evidence are supplied by a Knowledge Graph …
- Paradigm Shift in Quantitative Investment Driven by Large Language Models and Graph Spatiotemporal Networks: A Deep Empirical Study based on Market-Implied Sentiment and Neuro-Symbolic SystemsYi Wensheng · WSEAS TRANSACTIONS ON CIRCU... · Jul 20, 2026
Driven by the evolution of Large Language Models (LLMs), financial forecasting is shifting from statistical arbitrage to semantic intelligence. To address the limitations of traditional paradigms in processing unstructured data and modeling…
- DriftGuard-AEDL: Concept-Drift-Aware Continual Neuro-Symbolic Causal Inference for Evolving Time SeriesAnika Shah, Matteo Greco, Freya Nielsen · Journal of Computing and El... · Jul 20, 2026
Structural causal models estimated on non-stationary time series are threatened by concept drift: the data-generating structure itself changes over time, silently invalidating a once-correct model. Recent neuro-symbolic pipelines such as Ad…
- Differentiable Logic Programming to Mitigate Reasoning Shortcuts in Neurosymbolic SystemsAkihiro Takemura, Katsumi Inoue · Electronic Proceedings in T... · Jul 20, 2026
Neurosymbolic (NeSy) systems integrate neural networks with logical reasoning to achieve both generalization and interpretability, but recent work has shown they are susceptible to shortcut reasoning behaviors.We propose a novel method usin…
- From Industrial AI and the Mining AGI Platform to a Causal Neuro-Symbolic Mine-to-Market Framework: An Integrated Engineering Architecture for Mining Enterprises (Industrial Reference Edition)Durmishkhan Gakharia · Zenodo (CERN European Organ... · Jul 19, 2026
Mining AGI Platform: Industrial Reference Architecture and Implementation Package for AI-Enabled Mining Operations (Frozen pre-pilot baseline) This archive contains the complete engineering package accompanying the Mining AGI Platform conce…
- From Industrial AI and the Mining AGI Platform to a Causal Neuro-Symbolic Mine-to-Market Framework: An Integrated Engineering Architecture for Mining Enterprises (Industrial Reference Edition)Durmishkhan Gakharia · Zenodo (CERN European Organ... · Jul 19, 2026
Mining AGI Platform: Industrial Reference Architecture and Implementation Package for AI-Enabled Mining Operations (Frozen pre-pilot baseline) This archive contains the complete engineering package accompanying the Mining AGI Platform conce…
- Knowledge-Guided Neuro–Symbolic Modeling for Mineral Prospectivity Mapping: A Case Study of Porphyry Cu–Mo Systems in Boise County, IdahoWeilin Chen, Jiyin Zhang, Chenhao Li, Xiaogang Ma · Natural Resources Research · Jul 17, 2026
Abstract Mineral prospectivity mapping increasingly relies on machine learning to integrate geochemical and geological information, yet purely data-driven models often perform unreliably under sparse sampling, strong class imbalance, and co…
- Neuro-symbolic weak supervision: theory and semanticsNijesh Upreti, Vaishak Belle · Philosophical Transactions ... · Jul 16, 2026
Weak supervision enables machine learning models to learn from limited or noisy labels, but it introduces challenges in reliability and semantic clarity, particularly in multi-instance partial label learning (MI-PLL), where models must reso…
- A quantum enhanced neuro symbolic intrusion detection system for software defined networkingD. Lynda, G. Logeswari, K. Tamilarasi, G. Sudhakaran · Scientific Reports · Jul 16, 2026
Software Defined Networking (SDN) provides programmability and centralized control, but at the same time, increases the attack surface and prevents classical Intrusion Detection Systems (IDSs) from dealing with new attack vectors. To solve …
- From Maxwell to Analog AI: A Short Guide to Digital, Analog, Neuromorphic, Edge and Hybrid IntelligenceNermin Sökmen · Zenodo (CERN European Organ... · Jul 14, 2026
Artificial intelligence is often described through terms that belong to different classification layers. Digital and analog AI identify how computation is physically represented and executed; neuromorphic AI describes an architectural inspi…
- From Maxwell to Analog AI: A Short Guide to Digital, Analog, Neuromorphic, Edge and Hybrid IntelligenceNermin Sökmen · Zenodo (CERN European Organ... · Jul 14, 2026
Artificial intelligence is often described through terms that belong to different classification layers. Digital and analog AI identify how computation is physically represented and executed; neuromorphic AI describes an architectural inspi…
- The Neuro-Symbolic Singularity Decay Theorem: Aleph-Null Bypass EquationMuhammad Aidil Amry · Zenodo (CERN European Organ... · Jul 10, 2026
This research introduces a fundamental paradigm shift in automated theorem proving and Bounded Model Checking. Conventional Satisfiability Modulo Theories (SMT) solvers operate strictly within discrete Boolean spaces. When extracting Minima…
- The Neuro-Symbolic Singularity Decay Theorem: Aleph-Null Bypass EquationMuhammad Aidil Amry · Zenodo (CERN European Organ... · Jul 10, 2026
This research introduces a fundamental paradigm shift in automated theorem proving and Bounded Model Checking. Conventional Satisfiability Modulo Theories (SMT) solvers operate strictly within discrete Boolean spaces. When extracting Minima…
- Neuro-Symbolic pathways to AGI: compositional reasoning and trustworthy deploymentSafayat Bin Hakim, Kanchon Gharami, Huihui Wang, Muhammad Adil et al. · Progress in Artificial Inte... · Jul 10, 2026
Abstract Large-scale neural architectures exhibit systematic failures in compositional generalization and formal verifiability despite remarkable pattern recognition capabilities. This paper introduces the Neural-Symbolic-Verification (NSV)…
- Explainable multimodal AI and neuro-symbolic clinical decision support system for chronic eye disease management: a digital health implementation studyHan Wang, Simon Ming Yuen Lee, Guanghui Hou, Yapeng Wang et al. · Frontiers in Digital Health · Jul 9, 2026
Introduction Administrative burden and documentation workload are increasingly recognized as major contributors to healthcare costs and clinician burnout, particularly in chronic disease management such as age-related macular degeneration (…
- SkillCenter: A Large-Scale Source-Grounded Skill Library for Autonomous AI AgentsTianming Sha, Yue Zhao, Lichao Sun, Yushun Dong · arXiv · Jul 8, 2026
Autonomous AI agents can execute complex tasks with limited human review, yet they often lack the grounded operational knowledge to make their outputs not just executable but correct, secure, and maintainable. We introduce SkillCenter, to o…
- Accelerating NeurASP with Vectorization and CachingAlexander Philipp Rader, Alessandra Russo · Theory and Practice of Logi... · Jul 8, 2026
Abstract Neurosymbolic AI combines neural networks with symbolic programs to create robust and explainable predictions. One such framework is NeurASP, which trains a neural network to predict concepts and reasons over them using rules writt…
- G-RRM: Guiding Symbolic Solvers with Recurrent Reasoning ModelsTimo Bertram, Sidhant Bhavnani, Richard Freinschlag, Erich Kobler et al. · arXiv · Jul 2, 2026
In this work, we focus on SE-RRMs, a symbol-equivariant instantiation of RRMs that exhibits improved extrapolation to larger problem sizes. We propose a neuro-symbolic approach, ``Guiding with Recurrent Reasoning Models'' (G-RRM), which int…
- Grounding large language models in an anthropological knowledge graph: A neuro-symbolic approachMatt Artz · Human Organization · Jul 1, 2026
As artificial intelligence (AI) systems increasingly mediate access to knowledge, anthropologists face questions about how to engage with these technologies constructively. In this article I demonstrate how pairing large language models (LL…
- Memory-Efficient Probabilistic Neuro-Symbolic Integration for Explainable Natural Language Inference Using Transformer-Based Foundation ModelsZahraa Sameer Ibrahim, Haedar Ahmed Mukhef, Hayder Hasan Ali · Al-Mustansiriyah Journal of... · Jun 30, 2026
Background: Transformer-based foundation models have achieved state-of-the-art results in various natural language inference benchmarks, but their decision-making processes remain largely unexplainable. Addressing the ’explainability gap’ i…
- Neuro-Symbolic AI System for Logical Reasoning and Decision MakingSUSHMA S, J.LIN EBY CHANDRA · IJARCCE · Jun 28, 2026
The ABO-Rh blood group system constitutes a critical biomarker in transfusion medicine, organ transplantation, and prenatal care.Conventional serological determination methods require venipuncture, trained personnel, and laboratory infrastr…
- Deterministic Neuro-Symbolic Orchestration (DNSO): A Paradigm Shift in Autonomous AI ArchitecturesJason Gabriel Davis · Zenodo (CERN European Organ... · Jun 27, 2026
The current industry trajectory in Artificial Intelligence - specifically the pursuit of Autonomous General Intelligence (AGI) - is fundamentally bottle-necked by an architectural flaw: the attempt to use probabilistic Large Language Models…
- Deterministic Neuro-Symbolic Orchestration (DNSO): A Paradigm Shift in Autonomous AI ArchitecturesJason Gabriel Davis · Zenodo (CERN European Organ... · Jun 27, 2026
The current industry trajectory in Artificial Intelligence - specifically the pursuit of Autonomous General Intelligence (AGI) - is fundamentally bottle-necked by an architectural flaw: the attempt to use probabilistic Large Language Models…
- Deterministic Neuro-Symbolic Orchestration (DNSO): A Paradigm Shift in Autonomous AI Architectures Author/Inventor: Jason Gabriel Davis Date of Public Disclosure: June 26, 2026 Abstract The current industry trajectory in Artificial Intelligence—specifically the pursuit of Autonomous General Intelligence (AGI)—is fundamentally bottlenecked by an architectural flaw: the attempt to use probabilistic Large Language Models (LLMs) as both the reasoning engine and the memory store. This results in unavoidable systemic failures: stochastic hallucinations, semantic drift, memory contamination, and attention dilution. This disclosure introduces Deterministic Neuro-Symbolic Orchestration (DNSO), a novel architectural framework that solves these issues by decoupling logical reasoning from probabilistic synthesis. By strictly constraining an LLM to act solely as a tightly bounded text-processing subroutine—while a rigid, rule-based ontological schema acts as the sole epistemic authority—DNSO achieves autonomous, self-correcting cognitive loops without memory contamination. This framework has been successfully reduced to practice in two distinct proof-of-concept embodiments: KINTSUGI (applied real-time geopolitical and environmental predictive intelligence) and OLLIN (autonomous scientific hypothesis generation), both operating entirely on consumer-grade, offline hardware. 1. The Crisis of Probabilistic AI Contemporary AI systems attempt to build autonomous agents by wrapping LLMs in prompt chains. This fails for three deterministic reasons: Semantic Drift & Memory Bloat: When a probabilistic model hallucinates a pattern and writes it back to the system's memory, the epistemic foundation is permanently corrupted. Echo Chambers: Systems that use their own generated predictions as evidence for future predictions detach from empirical reality, creating runaway feedback loops. Attention Dilution: Forcing a model to evaluate massive context windows (e.g., 10,000-line logs) obscures high-signal structural anomalies in volumetric noise. 2. The DNSO Architectural Framework DNSO does not attempt to make an LLM reliable. Instead, it treats the LLM as a "mouth" (Wernicke's and Broca's areas)—a subordinate text-processing tool—while a deterministic Python/SQL pipeline acts as the "brain" (the prefrontal cortex and hippocampus). Core Architectural Rules: Epistemic Boundary Enforcement: A deterministic, immutable relational database is the sole location authority and epistemic gatekeeper. No event enters memory without deterministic validation. The "No Prediction as Evidence" Rule: Hypotheses (predictions) generated by the system are tracked as living objects but are architecturally barred from being used as "facts" to support subsequent predictions. Only finalized, empirically resolved outcomes are parsed back into the knowledge base. This mathematically neutralizes semantic drift. Continuous Self-Correction: The system autonomously monitors the outcome of its hypotheses against reality, identifies the specific flawed assumption (hinge point) that caused a failure, and dynamically updates its internal confidence weights without human intervention. 3. Embodiment I: KINTSUGI (Applied Predictive Intelligence) KINTSUGI is a real-time, applied intelligence architecture designed to autonomously ingest, classify, and predict cross-domain global events (geopolitical, environmental, economic). Novel Mechanisms: Autonomous Self-Critique Gate: Before a prediction is elevated to "Active" status, it must pass through a counterfactual reasoning gate. The system maintains a "Global Credibility Streak," requiring consecutive correct predictions before the self-critique gate relaxes. Failures increment a "recheck depth," forcing the system to scrutinize future predictions against the hinge points of past failures. Shannon Entropy OCB Margins: Anomaly detection is calculated using Operational Coherence Bound (OCB) margin mathematics, dynamically computing baselines for domain activity to determine elevated, high, or critical risk states. Dynamic Symbolic Weighting: The system learns by autonomously updating deterministic weight values in a relational database based on empirical resolution, while the generative LLM parameters remain static. 4. Embodiment II: OLLIN (Autonomous Scientific Discovery) OLLIN is a domain-agnostic research intelligence architecture applied to the field of autonomous scientific hypothesis generation. It proves the DNSO framework is universally applicable. Novel Mechanisms: Structural Isomorphism Mapping: OLLIN solves polymorphic evasion (changing bytecode/signatures) by hunting for structural isomorphisms rather than lexical matches. The retrieval layer penalizes lexical overlap and rewards cross-domain semantic distance, forcing the LLM to identify structural bridges between disparate fields. Constrained Sampling (Solving Attention Dilution): Instead of dumping massive context windows into an LLM, DNSO uses a "sweeJason Gabriel Davis · Zenodo (CERN European Organ... · Jun 27, 2026
To establish Prior Art...
- Comparing Emotional Signatures in Tamil Indie and Mainstream Songs Using Neuro Symbolic AIJurnal pengukuran kualiti d... · Jun 25, 2026
- NSM-AMG: Neuro-symbolic Multimodal Framework with Adaptive Modality Gating and ABCDE-aligned Reasoning for Skin Cancer DetectionRuchi Sharma, Dr. Archana Sandhu · International journal of in... · Jun 24, 2026
Reliable and interpretable skin cancer detection from dermoscopy images is an urgent clinical problem.Current deep learning methods largely treat this as a unimodal image classification problem, ignoring patient metadata, and produce opaque…
- Navigating in Latent Space and Retrieval-Augmented GenerationRobert Haase · Zenodo (CERN European Organ... · Jun 24, 2026
In this session of the ScaDS.AI Summer School 2026 on Neuro+Symbolic AI, we dive into vector embeddings using language models and vision transformers. We will use interactive tools for spatial exploration and we see how to use these embeddi…
- Navigating in Latent Space and Retrieval-Augmented GenerationRobert Haase · Zenodo (CERN European Organ... · Jun 24, 2026
In this session of the ScaDS.AI Summer School 2026 on Neuro+Symbolic AI, we dive into vector embeddings using language models and vision transformers. We will use interactive tools for spatial exploration and we see how to use these embeddi…
- Evolution of AI-Driven Game Space Generation: Deep Learning, Reinforcement Learning, and Neuro-Symbolic ArchitecturesYuehua Chen · Applied and Computational E... · Jun 23, 2026
In non-combative games (such as meat pigeon, life simulation and box puzzle games), the quality of architectural space directly affects the player's exploration process of the scene and the effect of environmental narrative. Traditional pro…
- Scaling Linear Mode Connectivity and Merging to Billion Parameter Pretrained TransformersTianyi Li, Zhiqiang Shen · arXiv · Jun 22, 2026
Linear mode connectivity (LMC) provides a promising foundation for understanding and merging independently trained neural networks, but existing methods typically optimize the interpolation path from only one model endpoint, limiting their …