hallucination mitigation

Hallucination remains a critical reliability challenge in the practical deployment of large language models (LLMs). Method: This work introduces, for the first time, a dual-dimensional taxonomy of hallucinations—distinguishing *knowledge-based* and *logic-based* types—and proposes a unified framework integrating retrieval-augmented generation (RAG), chain-of-thought (CoT) reinforcement, and agent-based system orchestration to systematically mitigate them. We analyze the intrinsic mechanisms by which each component suppresses distinct hallucination categories. Contribution/Results: Through rigorous empirical evaluation on standardized benchmarks, we systematically characterize the suppression pathways of each technique across hallucination types. The study delivers a reusable, modular paradigm for enhancing LLM reliability and a standardized evaluation framework. Our approach significantly improves both factual accuracy and operational feasibility—bridging the gap between theoretical robustness and real-world deployment.

Foundation models frequently exhibit “hallucinations” during autonomous decision-making, leading to high-risk misjudgments—yet no formal definition or systematic characterization of hallucination exists for decision tasks. Method: This work introduces the first task-specific definition of hallucination in decision-making and establishes a cross-task, scalable hallucination taxonomy; proposes a synergistic framework integrating uncertainty quantification with hallucination detection; and develops a joint detection–decision evaluation paradigm grounded in systematic survey analysis, probabilistic uncertainty modeling, and decision-chain interpretability. Results: We present the first comprehensive landscape of hallucination detection techniques tailored to decision contexts, contributing seven actionable implementation guidelines and identifying five critical research gaps—thereby advancing the safe deployment of trustworthy foundation models in high-stakes domains such as healthcare and transportation.

This work addresses the susceptibility of large language models to hallucinations in high-stakes domains such as finance and law, which undermines output reliability. The authors propose a root-cause-aware continuous improvement framework that categorizes hallucination sources into three types: model-induced, data-related, and context-driven. By integrating techniques including uncertainty estimation, reasoning consistency analysis, knowledge anchoring, and confidence calibration, the framework establishes a closed-loop mechanism for hierarchical detection and targeted mitigation. This approach shifts the paradigm from generic post-hoc fixes to precise, cause-specific governance. Evaluated on financial data extraction tasks, the method significantly enhances both generation accuracy and trustworthiness, offering a scalable solution for deploying reliable AI systems in regulation-sensitive scenarios.

Large Vision-Language Models (LVLMs) commonly suffer from hallucination—specifically, image-text response inconsistency. Existing instruction-tuning methods rely on generic high-quality datasets and overlook the model-specific distribution of hallucination concepts across different LVLMs, limiting their mitigation efficacy. This work is the first to reveal and quantify the concept-level model specificity of hallucinations in LVLMs. Based on this insight, we propose Directed Fine-Tuning Generation (DFTG), a two-stage framework: (1) multimodal response analysis and image-text consistency diagnosis to precisely identify the target model’s hallucination patterns; and (2) controllable synthesis of targeted instruction data guided by the diagnosis, followed by lightweight fine-tuning. Experiments demonstrate that DFTG significantly outperforms baselines such as LRV-Instruction across multiple hallucination evaluation benchmarks, substantially reducing hallucination rates and improving cross-modal faithfulness.

Large language models (LLMs) exhibit heterogeneous hallucinations, yet prior work lacks a principled taxonomy distinguishing their root causes. Method: This paper introduces the first systematic dichotomy: hallucinations arising from genuine knowledge gaps (HK−) versus those stemming from erroneous activation of existing knowledge (HK+). Through multi-model, cross-dataset experiments—augmented by knowledge probing and human-annotated causal attribution—we empirically establish HK+ as pervasive and highly model-specific. Building on this insight, we propose a novel paradigm for constructing model-specific hallucination datasets and train a lightweight detector to significantly improve fine-grained hallucination classification and detection accuracy. Contribution/Results: We release the first open-source toolkit supporting HK−/HK+ fine-grained annotation, enabling precise hallucination analysis and targeted mitigation. This work establishes a new diagnostic and intervention framework for LLM hallucinations grounded in causal mechanism disentanglement.

This work addresses the challenge of prompt-induced hallucinations in large language models by proposing a structured reasoning approach that integrates code-guided exploration with knowledge graphs. The method embeds executable code as a structured knowledge carrier within chain-of-thought prompting, enabling the model to perform controllable traversal over a knowledge graph. A chained knowledge distillation architecture further refines the reasoning process. This approach achieves the first implementation of code-driven external knowledge injection, significantly enhancing factual accuracy on both GPT-4 and LLaMA-3.3. Experimental results demonstrate consistent improvements across multiple settings, with HIT@1, HIT@3, and HIT@5 scores increasing by 15.64%, 13.38%, and 13.28%, respectively, and overall hit rates exceeding 95% in various evaluation scenarios.

To address hallucination in large vision-language models (LVLMs) during image captioning—caused by image-text misalignment—this paper proposes a fine-grained, AI-feedback-driven detection and mitigation framework. Methodologically, it introduces the first sentence-level, multi-type hallucination detector identifying object-, attribute-, and relation-level inconsistencies; designs hallucination-severity-aware direct preference optimization (HSA-DPO) to close a detection–rewriting–preference-learning loop; and operates entirely without human annotations or reliance on black-box foundation models. The key contribution is a lightweight, transferable end-to-end solution that significantly improves LVLM reliability across multiple benchmarks: hallucination detection F1 score increases by 12.6%, and image-text alignment of generated captions improves by 23.4%. This work establishes a novel paradigm for enhancing LVLM robustness and factual consistency.

This work addresses the critical challenge of hallucination-induced cascading failures in GUI agents during real-world deployment, where existing approaches lack fine-grained diagnosis, reliable evaluation, and efficient mitigation mechanisms. To this end, the paper proposes the first hallucination governance framework tailored for GUI environments. It introduces an empirically grounded hallucination taxonomy, a three-stage calibration and evaluation pipeline, and a lightweight closed-loop structured reasoning module augmented with a cold-start post-training strategy. Remarkably, the approach achieves a significant reduction in hallucination rates using only 9K training samples, substantially enhancing the agent’s environmental grounding and operational fidelity without requiring extensive computational resources.

This study addresses the critical issue of hallucinations in medical large language models (LLMs) when processing electronic health records, which poses significant risks to clinical decision-making. To tackle this challenge, the authors construct the first medical hallucination annotation dataset based on MIMIC-IV discharge summaries, incorporating dual verification by both medical students and GPT-4o. They further propose a parameter-free hallucination mitigation framework that integrates a self-reflection error-correction mechanism with retrieval-augmented in-context learning (RA-ICL). Experimental results demonstrate that this approach significantly improves generation accuracy (p<0.05) across three out of five open-source LLMs evaluated. The work releases a high-quality hallucination dataset and a reproducible framework, establishing a new paradigm for the safe deployment of medical LLMs.

This paper systematically surveys hallucination in large language models (LLMs), clarifying its root causes—including training data bias, decoding-induced distortion, and incomplete knowledge representation—and classifying its multidimensional manifestations (e.g., factual, logical, and consistency hallucinations). Methodologically, it proposes a unified hallucination taxonomy and establishes a three-layer mitigation framework—detection, attribution, and suppression—that integrates major technical paradigms: pretraining interventions, prompt engineering, retrieval-augmented generation, and post-hoc correction. Through structured literature analysis and mechanistic dissection of representative approaches, the work delivers the first practice-oriented, one-stop knowledge portal for hallucination governance. The resulting concise survey document enhances research reproducibility and accelerates real-world deployment, providing both theoretical foundations and methodological guidance for advancing LLM trustworthiness.

In enterprise settings, large language models (LLMs) suffer from hallucination due to limited context windows and outdated knowledge; existing mitigation strategies—such as gold-standard QA repositories or secondary verification models—are costly and lack formal guarantees of correctness. This paper proposes an interactive, visualization-enabled knowledge graph framework for hallucination detection: LLM-generated assertions are dynamically linked to proprietary knowledge sources to construct a structured truth-graph, supporting confidence scoring, provenance tracing, and human-in-the-loop feedback. Our key contributions lie in the integration of adaptive knowledge graph construction, interpretable natural language understanding (NLU), and human–AI collaborative diagnosis—enabling real-time identification and auditable verification of hallucinated content. Experiments demonstrate significant improvements in LLM response trustworthiness and reliability under constrained context and knowledge inconsistency, while establishing a sustainable, feedback-driven optimization loop.

This work addresses the tendency of multimodal reasoning models to over-rely on linguistic priors and neglect visual inputs when equipped with reasoning mechanisms, thereby exacerbating hallucination. To mitigate this issue, the authors propose C3PO, a novel framework that uniquely integrates chain-of-thought compression with inductive contrastive preference learning. By compressing redundant reasoning tokens to preserve essential visual information and leveraging high-quality AI-generated feedback to construct contrastive preference signals, C3PO explicitly suppresses multimodal hallucinations. The approach establishes a theoretically grounded training paradigm and demonstrates consistent and significant hallucination reduction across multiple state-of-the-art multimodal models and benchmark datasets.