Generative AI models (e.g., LLMs, diffusion models) are increasingly used to create multimodal content (text + images) for applications like interactive storytelling, educational tools, and digital media. However, ensuring that generated content is safe, unbiased, and age-appropriate remains a critical challenge. Manual moderation is unscalable, and existing automated filters often lack contextual understanding or multimodal reasoning.

 This thesis explores the development of automated pipelines to evaluate and refine AI-generated content, with a focus on:

• Real-time safety assessment of text-image pairs.
• Automatic re-prompting to guide models toward compliant outputs.
• Adaptability to diverse use cases (e.g., children’s toys, educational platforms).

Objectives

1. Multimodal Safety Evaluation:

   • Investigate state-of-the-art metrics (e.g., CLIP-based similarity, toxicity scores, emotional valence) to detect unsafe or age-inappropriate content in text-image pairs.
   • Develop a lightweight, interpretable scoring system combining:
     • Text analysis (e.g., perspective API, custom fine-tuned classifiers).
     • Image analysis (e.g., NSFW detectors, aesthetic/emotional classifiers).
     • Cross-modal alignment (e.g., does the image match the text’s intent and safety constraints?).

2. Re-Prompting Strategies:

   • Design adaptive prompting techniques to iteratively refine outputs (e.g., using reinforcement learning or constrained decoding).
   • Explore few-shot learning to generalize safety rules across domains (e.g., fairy tales vs. scientific explanations).

3. Benchmarking and Evaluation:

   • Curate a multimodal dataset of edge cases (e.g., subtle biases, ambiguous contexts).
   • Compare against human annotations and existing tools (e.g., Google’s Perspective API, LAION filters).
   • Optimize for latency and computational efficiency (critical for embedded/edge devices).

The Chair of Media Technology at the Technical University of Munich (TUM) is offering an exciting Master’s Thesis opportunity in the context of future 6G networks and digital twin-based optimization. The thesis will contribute to a cutting-edge research project focused on the planning and simulation of 6G wireless systems in complex environments.

The topic is split into the following steps.

Selection and Evaluation of Simulation software

As 6G networks are still under development, simulation is essential for validating novel methods. The first part of this thesis involves identifying and evaluating suitable 6G simulation platforms that can replicate key network functions and ray-tracing-based propagation models. The selected simulator should ideally support:

• 3D point clouds or triangular mesh input

• CAD data processing

• Modifiable object classes

• Material-aware modeling

Candidate simulators:

• DeepMIMO, https://www.deepmimo.net/, Letzter Zugriff Februar 2025
• NIST, https://github.com/usnistgov/qd-realization, Letzter Zugriff Februar 2025
• Hoydis, Jakob, et al. "Sionna: An open-source library for next-generation physical layer research." arXiv preprint arXiv:2203.11854 (2022).

Offline Optimization for 6G Network Deployment

Building upon the simulator selected above, the second part focuses on offline optimization of access point placement and beamforming strategies using a digital twin of the environment. This involves:

• Differentiating between static and dynamic objects (e.g., fixed machinery or columns)

• Incorporating material-specific signal propagation effects

• Developing methods to maximize signal coverage by optimizing AP locations in a virtual replica of the environment

Prerequisites