Presenting our RAID 2026 paper at the iCSS Seminar Series
On 20 October 2026, I gave a talk at the Institute of Cyber Security for Society (iCSS) Seminar Series at the University of Kent, on our RAID 2026 paper “Small, Free, and Effective: Orchestrating Open-Weight Small Language Models to Outperform Single LLM for Malware Analysis”. The seminar was held in a hybrid format, in the Kennedy Building and on MS Teams.
I shared how orchestrated ensembles of small, open-weight language models can match or exceed single LLMs on questions about malware detonation reports. Our best hybrid system, combining evidence-grounded pipelines with adversarial debate, outperformed both the strongest cyber-specialised model and the strongest ungrounded frontier LLM.
This work was done with my supervisors Dr. Budi Arief and Prof. Shujun Li.