Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hallucination Mitigation using Agentic AI Natural Language-Based Frameworks

Diego Gosmar, Deborah A. Dahl

TL;DR

This work tackles the persistent problem of hallucinations in large language models by deploying a structured, multi-agent AI framework that communicates via the OVON Open Voice Network API. A front-end agent generates responses to 310 deliberately crafted prompts designed to induce hallucinations, which are then sequentially reviewed and refined by second- and third-level agents, with a fourth KPI-evaluator agent quantifying improvements. The approach introduces a set of novel hallucin activity metrics—Factual Claim Density, Factual Grounding References, Fictional Disclaimer Frequency, and Explicit Contextualization Score—aggregated into the Total Hallucination Score to track progress. Empirical results show systematic reductions in THS across agent levels and demonstrate the value of structured NLP-based inter-agent communication for improving content reliability, with a concrete use-case illustrating practical application. The findings support the potential of agentic AI and OVON-based pipelines to advance Trustworthy AI, while highlighting areas for broader model diversity and enhanced explainability.

Abstract

Hallucinations remain a significant challenge in current Generative AI models, undermining trust in AI systems and their reliability. This study investigates how orchestrating multiple specialized Artificial Intelligent Agents can help mitigate such hallucinations, with a focus on systems leveraging Natural Language Processing (NLP) to facilitate seamless agent interactions. To achieve this, we design a pipeline that introduces over three hundred prompts, purposefully crafted to induce hallucinations, into a front-end agent. The outputs are then systematically reviewed and refined by second- and third-level agents, each employing distinct large language models and tailored strategies to detect unverified claims, incorporate explicit disclaimers, and clarify speculative content. Additionally, we introduce a set of novel Key Performance Indicators (KPIs) specifically designed to evaluate hallucination score levels. A dedicated fourth-level AI agent is employed to evaluate these KPIs, providing detailed assessments and ensuring accurate quantification of shifts in hallucination-related behaviors. A core component of this investigation is the use of the OVON (Open Voice Network) framework, which relies on universal NLP-based interfaces to transfer contextual information among agents. Through structured JSON messages, each agent communicates its assessment of the hallucination likelihood and the reasons underlying questionable content, thereby enabling the subsequent stage to refine the text without losing context. The results demonstrate that employing multiple specialized agents capable of interoperating with each other through NLP-based agentic frameworks can yield promising outcomes in hallucination mitigation, ultimately bolstering trust within the AI community.

Hallucination Mitigation using Agentic AI Natural Language-Based Frameworks

TL;DR

This work tackles the persistent problem of hallucinations in large language models by deploying a structured, multi-agent AI framework that communicates via the OVON Open Voice Network API. A front-end agent generates responses to 310 deliberately crafted prompts designed to induce hallucinations, which are then sequentially reviewed and refined by second- and third-level agents, with a fourth KPI-evaluator agent quantifying improvements. The approach introduces a set of novel hallucin activity metrics—Factual Claim Density, Factual Grounding References, Fictional Disclaimer Frequency, and Explicit Contextualization Score—aggregated into the Total Hallucination Score to track progress. Empirical results show systematic reductions in THS across agent levels and demonstrate the value of structured NLP-based inter-agent communication for improving content reliability, with a concrete use-case illustrating practical application. The findings support the potential of agentic AI and OVON-based pipelines to advance Trustworthy AI, while highlighting areas for broader model diversity and enhanced explainability.

Abstract

Hallucinations remain a significant challenge in current Generative AI models, undermining trust in AI systems and their reliability. This study investigates how orchestrating multiple specialized Artificial Intelligent Agents can help mitigate such hallucinations, with a focus on systems leveraging Natural Language Processing (NLP) to facilitate seamless agent interactions. To achieve this, we design a pipeline that introduces over three hundred prompts, purposefully crafted to induce hallucinations, into a front-end agent. The outputs are then systematically reviewed and refined by second- and third-level agents, each employing distinct large language models and tailored strategies to detect unverified claims, incorporate explicit disclaimers, and clarify speculative content. Additionally, we introduce a set of novel Key Performance Indicators (KPIs) specifically designed to evaluate hallucination score levels. A dedicated fourth-level AI agent is employed to evaluate these KPIs, providing detailed assessments and ensuring accurate quantification of shifts in hallucination-related behaviors. A core component of this investigation is the use of the OVON (Open Voice Network) framework, which relies on universal NLP-based interfaces to transfer contextual information among agents. Through structured JSON messages, each agent communicates its assessment of the hallucination likelihood and the reasons underlying questionable content, thereby enabling the subsequent stage to refine the text without losing context. The results demonstrate that employing multiple specialized agents capable of interoperating with each other through NLP-based agentic frameworks can yield promising outcomes in hallucination mitigation, ultimately bolstering trust within the AI community.
Paper Structure (14 sections, 2 equations, 6 figures, 2 tables)

This paper contains 14 sections, 2 equations, 6 figures, 2 tables.

Table of Contents

  1. Structure of the Paper
  2. Introduction
  3. Inducing Hallucinations
  4. Agentic AI and Multi-Agent Orchestration
  5. The OVON Universal API approach
  6. Experimental Design and Methods
  7. Hallucination KPIs
  8. Empirical Results
  9. Use case
  10. Result discussion
  11. Limitations
  12. Future Improvements
  13. Conclusion
  14. Acknowledgments

Figures (6)

  • Figure 1: Multi-Agent Scenario for Hallucination Mitigation Evaluation
  • Figure 2: THS results over 310 prompts with 3 agents
  • Figure 3: Delta THS hallucination mitigation by prompt
  • Figure 4: Total THS per agent
  • Figure 5: Inter-Agent Hallucination Mitigation
  • ...and 1 more figures