Advancing Semantic Anomaly Detection through Fine-Tuning of Large Language Models (LLMs)

Institut

Professur für autonome Fahrzeugsysteme

Typ

Semesterarbeit / Masterarbeit /

Inhalt

 

Beschreibung

Background: The field of robotics and autonomous systems is experiencing rapid growth, necessitating the ability to operate in complex and unpredictable environments. This advancement, however, is accompanied by the increased likelihood of encountering unforeseen anomalies that could precipitate significant system failures. Such anomalies, exemplified by incidents in autonomous vehicles like Tesla's autopilot disengagement or phantom braking due to visual misinterpretations, underscore a critical need. These are not mere failures of individual system components but rather indicate a broader lack of semantic understanding, illustrating the imperative for enhanced anomaly detection capabilities that mirror human-level reasoning.

Research Focus: This project is dedicated to investigating the application of fine-tuning techniques on large language models (LLMs) to bolster semantic anomaly detection within vision-based autonomous systems. Fine-tuning LLMs presents a viable strategy for improving their ability to discern and interpret complex scenarios, closely mimicking human cognitive processes. Through this research, we aim to develop an advanced monitoring framework that adeptly identifies semantic anomalies, thereby ensuring higher levels of safety and reliability in autonomous operations.

Objective: The appointed master's student will play a crucial role in pioneering the exploration of fine-tuning LLMs for enhanced semantic anomaly detection. Key responsibilities will include: Architecting and deploying a fine-tuned LLM-based monitoring framework tailored for scenarios involving autonomous driving. Executing comprehensive experiments to assess the efficacy of the fine-tuned LLM approach in anomaly detection and its congruence with human evaluative standards. Analyzing the performance, strengths, and potential limitations of the fine-tuned models, with an eye towards future research pathways that could further refine and expand the application of LLMs in semantic anomaly detection.

Offer: We are excited to offer a master’s position at our institution, inviting you to contribute to the forefront of research on semantic anomaly detection through the innovative fine-tuning of LLMs. This project not only allows you to engage with advanced AI and machine learning techniques but also places you at the cutting edge of addressing real-world challenges in the domain of autonomous systems.

 

Your Benefits (What we offer):

• Participation in groundbreaking research with tangible applications in the burgeoning field of robotics and autonomous systems.
• Acquisition of in-depth knowledge and skills in state-of-the-art AI techniques, particularly the fine-tuning of large language models.
• Experience with leading-edge technology and adherence to best practices in software development.
• The potential for contributing to scholarly articles in esteemed scientific publications

Voraussetzungen

Requirements (What you should bring with you):

• Proactive initiative and a creative, problem-solving mindset.
• Proficiency in English and programming, especially in Python or C++.
• A strong foundational understanding of Machine Learning principles.
• A keen interest in, or prior experience with, autonomous vehicles will be considered an advantage. Experience with common software development tools (e.g., Git, Ubuntu) is desirable.

Tags

AVS Brusnicki

Möglicher Beginn

sofort

Kontakt

Roberto Brusnicki
roberto.brusnickitum.de