Large language models (LLMs) can generate fluent diagnostic explanations that appear coherent while containing unsupported claims, omitted premises, or logically invalid inferences. Such hallucinations are especially dangerous in human–machine collaborative environments, where LLM outputs may influence clinical judgment, escalation decisions, and trust calibration. This paper introduces ValidLLP4LLM, a hallucination detection and reasoning-validation framework based on a Labeled Logic Program architecture for physician–LLM joint diagnosis. The framework integrates several complementary reasoning paradigms, including logic programming, argumentation, probabilistic inference, and abductive explanation. Within this architecture, abduction plays a central role: it generates candidate explanatory hypotheses for observed symptoms or claims, while the LLP layer tests whether these hypotheses are supported by evidence, consistent with domain constraints, resistant to counterarguments, and clinically admissible. Unlike single-formalism validators, ValidLLP4LLM enriches symbolic facts and rules with semantic, epistemic, temporal, provenance, defeasibility, and discourse labels. This allows the system to represent not only what is asserted, but also who asserted it, how strongly it is supported, when it applies, whether it is defeasible, and whether it occurs in a central or peripheral part of the diagnostic discourse. By incorporating discourse-aware weighting, the framework gives greater verification priority to nucleus claims while attenuating satellite or contextual statements. Experiments on benchmark hallucination-detection datasets show that LLP-based validation improves over classical symbolic validators, especially when combined with discourse modeling, while results on a challenging clinical narrative dataset highlight the difficulty of verifying implicit, ambiguous, and underspecified patient language. A human evaluation further indicates that logic-assisted explanations improve hallucination detection accuracy, interpretability, and user trust. Overall, the results suggest that discourse-aware labeled symbolic reasoning provides a robust and interpretable foundation for validating LLM reasoning in safety-critical human–AI diagnostic workflows.

We present an Information-Theoretic Model of Abduction for Detecting Hallucinations in Generative Models, a neuro-symbolic framework that combines entropy-based inference with abductive reasoning to identify unsupported or contradictory content in large language model outputs. Our approach treats hallucination detection as a dual optimization problem: minimizing the information gain between source-conditioned and response-conditioned belief distributions, while simultaneously selecting the minimal abductive hypothesis capable of explaining discourse-salient claims. By incorporating discourse structure through RST-derived EDU weighting, the model distinguishes legitimate abductive elaborations from claims that cannot be justified under any computationally plausible hypothesis. Experimental evaluation across medical, factual QA, and multi-hop reasoning datasets demonstrates that the proposed method outperforms state-of-the-art neural and symbolic baselines in both accuracy and interpretability. Qualitative analysis further shows that the framework successfully exposes plausible-sounding but abductively unsupported model errors, including real hallucinations generated by GPT-5.1. Together, these results indicate that integrating Information-Theoretic divergence and abductive explanation provides a principled and effective foundation for robust hallucination detection in generative systems.

Large Language Models (LLMs) are increasingly used in industrial monitoring and decision support, yet they remain prone to process-control hallucinations-diagnoses and explanations that sound plausible but conflict with physical constraints, sensor data, or plant dynamics. This paper investigates hallucination as a failure of abductive reasoning, where missing premises, weak mechanistic support, or counter-evidence lead an LLM to propose incorrect causal narratives for faults such as pump restriction, valve stiction, fouling, or reactor runaway. We develop a neuro-symbolic framework in which Abductive Logic Programming (ALP) evaluates the coherence of model-generated explanations, counter-abduction generates rival hypotheses that test whether the explanation can be defeated, and Discourse-weighted ALP (D-ALP) incorporates nucleus-satellite structure from operator notes and alarm logs to weight competing explanations. Using our 500-scenario Process-Control Hallucination Dataset, we assess LLM reasoning across mechanistic, evidential, and contrastive dimensions. Results show that abductive and counter-abductive operators substantially reduce explanation-level hallucinations and improve alignment with physical process behavior, particularly in "easy-but-wrong'' cases where a superficially attractive explanation contradicts historian trends or counter-evidence. These findings demonstrate that abductive reasoning provides a practical and verifiable foundation for improving LLM reliability in safety-critical process-control environments.

This paper introduces a novel computational framework for analyzing and supporting invention dialogues, focusing on the interplay between problem-solving and creative reasoning. We propose a formalism for invention dialogues that captures the structure, patterns, and phase transitions. We leverage category-theoretic representation to describe the transformations in idea development, effectively modeling the processes of abstraction, refinement, and integration. Third, we frame invention as a Constraint Satisfaction Problem (CSP), enabling the representation of constraints and dependencies within the design process. Also, we present a Unified Model for Creative Reasoning, integrating dialogue formalism, category theory, and CSP frameworks into a cohesive computational approach. Finally, we explore the pedagogical value of invention dialogues, demonstrating their potential for training and enhancing creative problem-solving skills. Our contributions provide a robust theoretical foundation and practical tools for understanding and fostering innovation in both human and machine contexts in the form of invention dialogues. We evaluate the pedagogical value of automatically constructed invention dialogue and observe a substantial improvement of learning efficiency in comparison to plain text form of learning material.

The Social Promotion Copilot (SPC) is designed to autonomously engage with users on social platforms, leveraging advanced NLP and action-oriented automation. This chapter delves into its architectural foundation, focusing on the integration of Theory of Mind to enhance SPC’s ability to interpret user emotions, predict reactions, and tailor responses dynamically. By modeling mental states, SPC can distinguish between different user intents, such as seeking support, expressing dissatisfaction, or driving discussions. This capability enables more context-aware and persuasive interactions, making SPC a more effective tool for both social engagement and marketing strategies. Beyond cognitive modeling, the chapter also examines the run-time execution framework, detailing how SPC processes textual tasks in real-time. The system selects optimal actions based on multi-modal inputs, including textual context, sentiment analysis, and platform-specific engagement patterns. Through a structured decision-making pipeline, SPC adapts its posting and response strategies to maximize visibility and interaction. By balancing automation with adaptive intelligence, SPC transforms from a simple content-promotion tool into an autonomous social agent, capable of managing long-term engagement and fostering meaningful digital interactions, which is confirmed by evaluation on social network sites.

Large language models (LLMs) frequently generate fluent chain-of-thought (CoT) reasoning that appears coherent while containing unsupported inferences, omitted alternatives, or logically invalid conclusions, creating significant challenges for trustworthy and explainable AI in high-stakes domains such as healthcare. While prior work has primarily focused on factual hallucinations, the structural characteristics of hallucinated reasoning remain insufficiently understood. This paper investigates diagnostic hallucinations through the lens of discourse structure and introduces a synthetic benchmark of ambiguous patient complaints paired with diagnoses, reasoning traces, discourse-tree representations, and hallucination labels. We hypothesize that hallucinated CoT differs from grounded reasoning not only in factual correctness but also in discourse organization. Our analysis shows that hallucinated reasoning tends to elevate weak or speculative clues into central discourse nuclei, suppress contradictory evidence, and use discourse satellites for post hoc reinterpretation rather than evidence integration. In contrast, grounded reasoning preserves alternative hypotheses, maintains explicit contrast and concession relations, and resolves evidential conflicts at higher discourse levels. Based on these observations, we formulate discourse-level indicators for hallucination detection and reasoning verification. The resulting framework provides interpretable structural explanations of reasoning failures and supports neuro-symbolic validation of LLM-generated reasoning traces, contributing to the development of more trustworthy, explainable, and verifiable language models.

Large language models (LLMs) frequently generate coherent-looking reasoning traces that nevertheless contain unsupported inferences, contradiction suppression, premature closure, or structurally unstable explanations. Existing hallucination detection approaches largely focus on token-level uncertainty, factual verification, or post-hoc saliency analysis, while insufficiently modeling the discourse organization of reasoning itself. This paper proposes a discourse-aware extension of Concept Whitening (CW), originally developed for interpretable image recognition, to the problem of hallucination detection in chain-of-thought reasoning. We argue that hallucinations correspond to recurring structural distortions in discourse trees and rhetorical organization. Our framework aligns latent dimensions of neural reasoning representations with interpretable discourse concepts such as contradiction omission, unsupported nucleus promotion, defeater suppression, and abductive instability. By combining discourse parsing, rhetorical structure theory (RST), graph-based reasoning representations, and latent-space concept alignment, the proposed framework transforms hallucination detection into an interpretable geometric reasoning problem. We further analyze the computational properties of discourse-aware concept whitening and discuss its implications for neuro-symbolic reasoning verification.

A text obtained by a Large Language Model (LLM) such as GPT4 usually has issues in terms of incorrectness and hallucinations. We build a fact-checking system 'Truth-O-Meter' which identifies wrong facts, comparing the generation results with the web and other sources of information, and suggests corrections. Text mining and web mining techniques are leveraged to identify correct corresponding sentences; also, the syntactic and semantic generalization procedure adopted to the content improvement task. To handle inconsistent sources while fact-checking, we rely on an argumentation analysis in the form of defeasible logic programming. We compare our fact checking engine with competitive approach based on reinforcement learning on top of LLM or token-based hallucination detection. It is observed that LLM content can be substantially improved for factual correctness and meaningfulness.

We review the literature on medical discourse and attempt to build a computational model of it. Medical discourse sheds a light on communication structure of patient-doctor and other communication scenarios in healthcare and should be leveraged to facilitate and automate this communication when it is possible and practical. We propose a unified framework to represent communication discourse at the meta-level, where the subject of the communication is expressed in a language object. So far, the broad range of work on medical discourse is detached from computational discourse analysis, and we explore the possibilities of filling this gap and computationally treat the peculiarities of how information is passed between the agents in a hospital setting. We select the domain of question answering (QA) against a corpus of medical documents of diverse nature to evaluate our computational model of medical discourse. It turns out that applying specific structures obtained in medical discourse studies improves the relevance and efficiency of question answering.

Large Language Models (LLM) often produce text with incorrect facts and hallucinations. To address this issue, we developed a fact-checking system Truth-O-Meter12 which verifies LLM results on the Internet and other sources of information to detect wrong claims/facts and proposes corrections for them. NLP and reasoning techniques such as Abstract Meaning Representation and syntactic alignment are applied to match hallucinating sentences with truthful ones. To handle inconsistent sources while fact-checking, we rely on argumentation analysis in the form of defeasible logic programming, selecting the most authoritative source. Our evaluation shows that LLM content can be substantially improved for factual correctness and meaningfulness on an industrial scale.

In spite of the great progress of large language models (LLMs) in recent years, there is a popular belief that their limitations need to be addressed “from outside”, by building hybrid neurosymbolic systems which add robustness, explainability, perplexity and verification done at a symbolic level. We propose shape-charged learning in the form of Meta-learning/DNN - kNN that enables the above features by integrating LMM with explainable nearest neighbor learning (kNN) to form the object-level, having deductive reasoning-based metalevel control learning processes, performing validation and correction of predictions in a way that is more interpretable by humans.

In spite of great progress in recent years, deep learning (DNN) and transformers have strong limitations for supporting human–machine teams due to a lack of explainability, information on what exactly was generalized, and machinery to be integrated with various reasoning techniques, and weak defense against possible adversarial attacks of opponent team members. Due to these shortcomings, stand-alone DNNs have limited support for human–machine teams. We propose a Meta-learning/DNN → kNN architecture that overcomes these limitations by integrating deep learning with explainable nearest neighbor learning (kNN) to form the object level, having a deductive reasoning-based meta-level control learning process, and performing validation and correction of predictions in a way that is more interpretable by peer team members. We address our proposal from structural and maximum entropy production perspectives.

We extend the concept of a discourse tree (DT) in the discourse representation of text towards data of various forms and natures. The communicative DT to include speech act theory, extended DT to ascend to the level of multiple documents, entity DT to track how discourse covers various entities were defined previously in computational linguistics, we now proceed to the next level of abstraction and formalize discourse of not only text and textual documents but also various kinds of accompanying data. We call such discourse representation Multimodal Discourse Trees (MMDTs). The rational for that is that the same rhetorical relations that hold between text fragments also hold between data values, sets and records, such as Reason, Cause, Enablement, Contrast, Temporal sequence. MMDTs are evaluated with respect to the accuracy of recognition of criminal cases when both text and data records are available. MMDTs are shown to contribute significantly to the recognition accuracy in cases where just keywords and syntactic signals are insufficient for classification and discourse-level information needs to be involved.

We address the problem of argument detection by investigating discourse and communicative text structure. A formal graph-based structure called communicative discourse tree (CDT) is used. It consists of a discourse tree (DT) with additional labels on edges, which stand for verbs. These verbs represent communicative actions. Discourse trees are based on rhetoric relations, extracted from a text according to Rhetoric Structure Theory. The problem is tackled as a binary classification task, where the positive class corresponds to texts with arguments and the negative class corresponds to texts with no argumentation. The feature engineering for the classification task is conducted, deciding which discourse and communicative features are better associated with argumentation. New Intense Argumentation dataset is built and described. Mixed dataset including different types of argumentation and different text genres is collected. Evaluation on this mixed dataset is provided.