News
For an AI to work reliably with a company's own data such as contracts, records, or policies, it needs a method that finds the relevant passages for each question and provides them to the model.
The problem: conventional methods select these passages based only on superficial similarity, often resulting in incomplete or bloated inputs.
Building on a novel tripartite knowledge graph, Perelyn has developed a method that instead anchors the search to clearly defined domain concepts, enabling it to generate shorter, more precise, and traceable inputs, which directly translates into fewer errors, lower costs, and verifiable sources.
The method was presented at the "Knowledge Management" workshop at KI 2025 in Potsdam and has now been published in the proceedings.
Research at heart. Business in mind.
Here the link to the proceedings.
Event
Liliya Imasheva presented a validation pipeline for evaluating AI summaries at Conf42 Large Language Models 2026.
Event
At the IOCMA 2026 conference, Perelyn presented our method, which transforms graph data such as supply chains or networks so that AI models can learn more reliably from it, even over long distances.
