Banner Banner

Research Assistant (E13, Agility Project: "FakeXplain")

The Berlin Institute for the Foundations of Learning and Data (BIFOLD) at TU Berlin (Prof. Klaus-Robert Müller / Prof. Volker Markl) is looking for a research assistant in the field of machine learning and computational linguistics for an agility subproject in the domain of Language Language Models (LLMs) and eXplainable Al for disinformation detection. Prof. Möller's working group is developing methods for generating natural language explanations of LLMs for intelligent decision support in a fact-checking context. The goal of the project "FakeXplain - Development of transparent and meaningful explanations in the disinformation detection context" is to develop different approaches to integrate explainability such as Chain-of-Thought Prompting and Mechanistic Interpretability that are both understandable and meaningful for human users of LLM-based intelligent decision support systems. In this project we expect independent and responsible research in the areas described but also the evaluation of alternative approaches. Teaching duties are performed.

Requirements

Salary grade: TV-L 13, Berliner Hochschulen
Starting date: Earliest possible (limited for 3 years)
Closing date: April 26, 2024
Full job posting: IV-124/24