Enhancing Semantic Anomaly Detection using Low-Rank Adaptation (LoRA) of LLMs

Institut

Professur für autonome Fahrzeugsysteme

Typ

Semesterarbeit / Masterarbeit /

Inhalt

 

Beschreibung

Background: In the rapidly evolving domain of robotics and autonomous systems, the capacity to navigate complex environments with high levels of sophistication is increasingly crucial. However, this advancement brings with it the challenge of unexpected anomalies that can lead to critical system failures. Examples of such anomalies include unexpected behaviors in autonomous vehicles, such as Tesla's autopilot disengagement or phantom braking triggered by misleading visual cues. These issues stem from a systemic deficiency in semantic understanding rather than the malfunctioning of individual components, highlighting the need for improved anomaly detection mechanisms that are capable of sophisticated reasoning.

Research Focus: This project seeks to explore the integration of Low-Rank Adaptation (LoRA) techniques with large language models (LLMs) to improve semantic anomaly detection in vision-based systems. LoRA offers a promising avenue for enhancing the adaptability and efficiency of LLMs in recognizing complex, nuanced scenarios without extensive retraining. By leveraging LoRA, this research aims to develop a refined monitoring framework that can accurately identify semantic anomalies, aligning closely with human reasoning processes.

Objective: The selected master’s student will be at the forefront of investigating how LoRA can be applied to foundation models to significantly improve semantic anomaly detection. This includes: Designing and implementing a LoRA-enhanced monitoring framework within the context of autonomous driving. Conducting rigorous experiments to evaluate the effectiveness of the LoRA-based approach in identifying semantic anomalies, with a focus on comparison against traditional methods. Providing a detailed analysis of the strengths and limitations of using LoRA for semantic anomaly detection and proposing future directions for research.

Offer: We invite applications for a master’s position at our university, where you will have the unique opportunity to contribute to cutting-edge research on semantic anomaly detection using LoRA. This project offers the chance to delve into advanced AI techniques and apply them to real-world challenges, paving the way for significant contributions to the field of autonomous systems.

 

Your Benefits (What we offer):

• Engage in groundbreaking research with practical applications in robotics and autonomous systems.
• Gain expertise in advanced AI techniques, including Low-Rank Adaptation and large language models.
• Work with state-of-the-art technology and software development practices.
• Opportunity for publication in reputable scientific journals.

Voraussetzungen

Requirements (What you should bring with you):

• Strong initiative and a creative approach to problem-solving.
• Excellent command of English and programming skills in Python or C++.
• Solid foundation in Machine Learning concepts.
• Interest or prior experience in autonomous vehicles is an asset.
• Familiarity with software development practices (Git, Ubuntu) is preferred.

Tags

AVS Brusnicki

Möglicher Beginn

sofort

Kontakt

Roberto Brusnicki
roberto.brusnicki@tum.de