The goal of the Gen Learn Center is to lead ground-breaking research into state-of-art Generative Machine Learning and help HES-SO collaborators, students, and partners to better understand and use them.
In addition to the direct research activities into LLMs, Gen Learn Center also provides a convergence point for HES-SO collaborators and partners interested in introducing LLMs in their research, teaching, or operations.
Given the novelty of widely-accessible LLMs, this domain is in rapid evolution, and we are looking for partners to help develop best practices for LLMs development and usage together.
Do not hesitate to reach out to us if you want to collaborate!
Our current axes of research are:
- Fundamental research into Generative ML models and LLMs:
- Trade-offs between privacy, robustness, generalization abilities, and data efficiency
- Formal ways of representing Generative Models' inputs and outputs
- LLMs interaction and coordination as growable modules of dynamic applications
- Generative Learning on cyber-security, cyber-defense, and healthcare:
- LLMs' vulnerability to Jailbreaking
- Vulnerabilities of code generated by LLMs
- Suitability of LLMs for information operations in Switzerland
- Refinement of LLMs for direct patient interaction
- Providing experimentation capability to professors and students:
- On-premises
- With models with well-understood capacities and failure modes.
- Validation and refinement of LLMs for:
- Adherence to social norms
- Alignment with values of Switzerland, HES-SO, and application domains
- Acceptable behaviors
- Specific Applications
- QA of third-party models for:
- Biases
- Sandbox mode escapes
- Training data used
- Refinement of LLMs for specific applications:
- Research
- Internal capabilities
- Teaching
Contact e-mails: