CS MSc Thesis Presentation 17 February 2026

Tuesday, 17 February there will be a master thesis presentation in Computer Science at Lund University, Faculty of Engineering.

The presentation will take place in E:4130 (Lucas).

Note to potential opponents: Register as an opponent to the presentation of your choice by sending an email to the examiner for that presentation (firstname [dot] lastname [at] cs [dot] lth [dot] se). Do not forget to specify the presentation you register for! Note that the number of opponents may be limited (often to two), so you might be forced to choose another presentation if you register too late. Registrations are individual, just as the oppositions are! More instructions for opponents are found here on the LTH thesis project page.

11:15-12:00 in E:4130 (Lucas)

• Presenters: Ebba Rakuljic Feldt, Elin Hellström
• Title: Knowledge Graph–enhanced RAG for Enterprise Question Answering systems
• Examiner: Jacek Malec
• Supervisors: Marcus Klang (LTH), Nathan Hoche (Axis Communications AB)

As new knowledge emerges rapidly, the knowledge base of large language models (LLMs) can quickly become outdated, requiring frequent retraining, an expensive and resource-intensive process. Retrieval-augmented generation (RAG) addresses this by providing LLMs with up-to-date or domain-specific knowledge at inference time. However, traditional RAG systems struggle with broad, summarizing queries and complex reasoning, since retrieval is primarily based on semantic similarity. Recently, an alternative approach has been proposed that combines RAG with a knowledge graph (GraphRAG) to improve logical consistency and reasoning. In this thesis, we compare the performance of GraphRAG and traditional RAG on both internal enterprise data from Axis and the GraphRAG-Bench benchmark. For automatic construction of the knowledge graph, we use LLMs to extract entities and relations. To evaluate the enterprise data, we generate a synthetic question–answer dataset using LLMs and use an LLM-as-a-judge to assess performance. Our results show that, under the evaluated conditions, traditional RAG achieves higher performance than GraphRAG on many QA tasks, both on the benchmark and in the LLM-based evaluation of internal data.

Link to popular science summary