Research

A framework for counterfactual audit analysis in multi-agent systems — identifying the minimal changes to evidence, contract clauses, or protocols that would have altered an agent's decision.

A practical pipeline for classifying SME bank transactions using synthetic data augmentation, FinBERT fine-tuning, and temperature-scaling calibration to handle business context specific transaction descriptions and improve automated categorisation for cash-flow lending models.

Scalable temporal network modeling framework measuring co-evolution of online engagement with misinformation and its visibility using 100 million pandemic-related retweets. The framework can also be used to study other large-scale events where online attention is at stake, such as technological disruptions.

A dynamic benchmarking framework to assess LLM-based conversational agents through interactions with synthetic users, evaluating information extraction, context awareness, and adaptive engagement.

Large Language Models transform automated categorisation of SME bank transactions, improving accuracy and robustness with unstructured data and industry-specific terminology.