AI TRENDS | 0G Labs and University Teams Introduce Agora Automated Testing Framework in ICML 2026 Paper

0G Labs and research teams from the National University of Singapore, Peking University, and Beijing University of Posts and Telecommunications proposed an automated testing framework called Agora in an ICML 2026 accepted paper. According to Odaily, the framework combines domain knowledge with large-model multi-agent collaboration, using three specialized agents—coordinator, strategist, and code agent—to drive an automated testing architecture. In tests on consensus protocol libraries including Raft, EPaxos, HotStuff, and BullShark, Agora identified 15 previously unknown protocol-level deep logic vulnerabilities. The report said native large models such as GPT-5.2 and Claude 4.5 did not find vulnerabilities in comparison. Evaluation results showed that 73.9% of the vulnerability reports produced by Agora were real logic vulnerabilities. On average, it consumed about 5.32 million tokens per logic vulnerability found, equivalent to about $40. The framework’s code has been released to the open-source community.