Arxiv Papers

Enabling Open-World Specification Mining via Unsupervised Learning

A framework for mining specifications and usage patterns without the aid of rule templates, user-directed feedback, or predefined API surfaces. This paper leverages both learned embeddings and traditional co-occurrence statistics to disentangle trace-based data.