Research Article
Classification and Scenarios Analysis of Hallucination Suppression Methods Based on ReAct Framework of Reflection Mechanism
@INPROCEEDINGS{10.4108/eai.22-5-2026.2365227, author={Ruiyi Wang}, title={Classification and Scenarios Analysis of Hallucination Suppression Methods Based on ReAct Framework of Reflection Mechanism}, proceedings={Proceedings of the 4th International Conference on Image, Algorithms, and Artificial Intelligence, ICIAAI 2026, 22-24 May 2026, Singapore, Singapore}, publisher={EAI}, proceedings_a={ICIAAI}, year={2026}, month={8}, keywords={ReAct framework reflection mechanism hallucination suppression method classification adaptation scenario analysis}, doi={10.4108/eai.22-5-2026.2365227} }- Ruiyi Wang
Year: 2026
Classification and Scenarios Analysis of Hallucination Suppression Methods Based on ReAct Framework of Reflection Mechanism
ICIAAI
EAI
DOI: 10.4108/eai.22-5-2026.2365227
Abstract
The illusion problem has become a major limitation that limits their reliability. The mechanism of reflection and ReAct framework give a valuable research path to enhance the reliability of model inference. The current study concentrates on illusion suppression in the system of ReAct in relation to the reflection mechanisms with the systematic review of the contemporary mainstream reflection techniques, alongside the comparative analysis of these techniques, in the light of illusion rate and accuracy. The findings demonstrate that the system of reflection could enhance the reliability of the inference and stability of the task execution in the ReAct system to a significant extent. Various reflection approaches have distinguished their merits in terms of performance enhancement and cost of computation where the hybrid reflection approach performs the most favorable in terms of combined performance when dealing with complex tasks.
