Agentic AI Ensemble Systems for Real-World Reliability (WANGW_U26CMP)

About the Project

While large language models (LLMs)-based autonomous agents have shown immense potential, their deployment in complex, real-world workflows is hampered by hallucinations, logical inconsistencies, and brittleness under unforeseen constraints. Single-agent systems often lack the self-correction mechanisms necessary for high-stakes environments where errors carry significant operational risks.

This research aims to develop a robust Multi-Agent Ensemble Framework to enhance accuracy and reliability through collaborative intelligence. Key objectives include:

• Developing adversarial debate protocols where specialized agents cross-verify outputs to mitigate hallucinations.
• Implementing dynamic role allocation to assign tasks to agents based on real-time performance metrics and domain expertise.
• Integrating human-in-the-loop (HITL) triggers that quantify uncertainty and solicit human intervention when confidence thresholds are not met.

The project investigates various agent architectures, incorporating diverse foundation models and tool-use capabilities and Reinforcement Learning from Operational Feedback (RLOF) for fine-tuning the ensemble’s coordination logic. Evaluation will focus on "stress-testing" the system against non-stationary, real-world datasets in such as healthcare, agriculture, environmental science, finance etc.

The project should deliver a scalable architecture for reliable autonomous agency, providing a blueprint for AI systems that are not just high-performing, but demonstrably dependable in unpredictable, mission-critical applications.

The School of Computing Sciences (https://www.uea.ac.uk/about/school-of-computing-sciences) provides a vibrant research environment for conducting Computing and allied research and training. We collaborate with multi-national companies such as Apple, BT, the National Trust and Aviva, research institutes in the Norwich Research Park (https://www.norwichresearchpark.com), as well as other universities and industries in the UK and overseas. We are also members of the Turing University Network, a group of 65 UK universities working together to advance world-class research and build skills for the future.

The successful candidate will also be expected to contribute to Tutor activities for laboratory support on our BSc and MSc Courses in Artificial Intelligence, Data Science, and Computing Sciences commensurate with their core expertise, within the working hours permitted for full-time Postgraduate Researchers.

Funding Notes

This PhD project is in a competition for a Faculty of Science funded studentship. Funding is available to UK applicants and comprises ‘home’ tuition fees, an annual stipend for 3 years and a research training support budget.