About the RTG "Neuroexplicit Models"

Cutting-edge research

Neural models have revolutionized artificial intelligence, but they also have systematic limitations.

The research in this RTG will help overcome these limitations by combining neural models with human-interpretable explicit models into neuroexplicit models.

Learn More 

World-class faculty

Each PhD student in the RTG is co-supervised by two Principal Investigators, often across research area boundaries.

These are some of the top researchers in their fields in Europe and worldwide.

Learn More 

In-depth qualification program

An RTG is not just a research project - it is a training program for PhD students. Students in the RTG are part of a vibrant community that meets for joint activities and collaborates intensively with each other.

Learn More 

Research Program

The goal of this Research Training Group is to develop novel neuroexplicit models that accurately solve tasks in natural language processing, computer vision, and action-decision making, and to investigate the theoretical and practical principles of designing effective neuroexplicit models. Up to 24 PhD students at a time, together with 13 PIs and ~20 associated PhD students and postdocs, will carry out research on neuroexplicit models at the highest international level.

Deep neural models have revolutionized artificial intelligence over the past ten years, by learning from data how to perform a variety of AI tasks at unparalleled accuracy. By contrast, explicit models capture knowledge about a task or a domain in a way that can be understood or authored by human experts, and can therefore be more interpretable and data-efficient. Explicit models can use symbolic representations, or they can capture domain knowledge in other ways, e.g. through differential equations that describe the physics of the world. Neuroexplicit (and neurosymbolic) models combine neural and explicit elements, inheriting the complementary strengths of both.

Neural models have produced quantum leaps in what artificial intelligence can achieve. However, they also have systematic limitations, especially with respect to generalization, robustness, and interpretability. We and others have demonstrated that neuroexplicit models have the potential to overcome these limitations. However, neuroexplicit models need to be designed with care, and the principles of effective design are not well understood. By investigating the rich and varied landscape of neuroexplicit models across multiple areas of artificial intelligence in one coherent, interdisciplinary research group, the RTG will boost our understanding of these models. The RTG is the first research center and in particular the first PhD training program for neuroexplicit methods in Europe.

PhD students in the RTG choose their thesis topic freely, together with their two advisors. We welcome research topics that cut across the boundaries of research fields as much as in-depth advances within a field; some examples are shown below. Each PhD student is funded for four years and has access to travel funds, compute resources, and organizational support. The admission process is selective - we are looking for excellent PhD students in the world who will produce impactful research and become leaders in academia or industry.

Research in the RTG is loosely organized into four research areas: Language, Vision, Action, and Foundations.

LANGUAGE

Area L

PIs: Demberg, Klakow, Koller, Toneva

Neural models of language, such as GPT-4, have shown groundbreaking performance across even very complex natural language processing tasks. However, these models are data-hungry and slow, intrinsically uninterpretable, struggle with hallucinations and certain kinds of generalization, and are vulnerable to bias and noise in the data. We aim to combine neural and symbolic models that are consistently truthful, can reason robustly, and are linguistically interpretable.


Example thesis topics:

- Algebras for trainable sentence embeddings
- Goal-driven text generation with latent-state planning
- Complex entities in the brain and in DNNs
- Symbolic supervision of neural learning
- Neuroexplicit learning of pragmatic implicatures
- Explaining and intervening biases in transformers
- Task-dependent compositionality in the brain

VISION

Area V

PIs: Ilg, Schiele, Theobalt

Neural models have enabled great improvements across a wide range of computer vision problems, such as estimating correspondences between images and reconstruction of humans and objects in their environment.. They are efficient and scale to diverse scenes, but they rely on large quantities of training data and are severely lacking in generalization, robustness, and interpretability. We aim to overcome these limitations by combining them with parametric models, e.g. of the world physics, and set new states of the art in accuracy and runtime efficiency.


Example thesis topics:
- Neuroexplicit incremental learning of non-stationary 3D object distributions
- Deep learned optical flow with explicit diffusion
- Pragmatically contrastive summarization for instructional videos
- Neuroexplicit models for monocular human modeling
- Interpretable object-centric future video prediction
- Neuroexplicit uncertainty for 3D reconstruction
- A joint model for spoken language and gestures

ACTION

Area A

PIs: Hoffmann, Slusallek, Wolf

Deep reinforcement learning models can beat human world champions in complex games such as chess and Go and are successful across many problems in sequential decision making. However, they do not generalize across related problems, struggle to handle events that were rare in training, and are hard to trust due to lack of interpretability. We aim to learn explicit abstractions that improve generalization and improve the robustness towards rare events and uncertain sensor values through explicit simulations of the environment.


Example thesis topics:
- Robust reinforcement learning in the presence of rare events
- Explicit abstractions over neural state embeddings
- Robust driving decisions via neuroexplicit POMDPs
- Neuroexplicit abstractions for deep reinforcement learning
- Improved sampling strategies for sparse environments

FOUNDATIONS

Area F

PIs: Valera, Vreeken, Weickert

The RTG aims not only to improve the state of the art in Language, Vision, and Action, but also to identify the underlying principles of developing effective neuroexplicit models. We aim to develop novel theory and methods in machine learning that directly improve generalization, robustness, and interpretability and cut across the different application areas. This includes training methods that generalize across different domains and objectives, as well as models that can be provably seen as either explicit or neural and thus will helps us overcome the distinction between the modeling paradigms.


Example thesis topics:
- Generalization by learning task-invariant representations
- Auxiliary tasks for robust sequential decision making
- Compactly describable neural networks
- Interpretable deep neural networks for computer vision
- Enforcing neural network stability during learning
- Stability measures for deep neural networks and limitations on the learnability of evolution equations

Qualification Program

A Research Training Group (RTG) is not just a research project - it is a training program for PhD students. Students in the RTG are part of a vibrant community that meets for joint activities and collaborates intensively with each other.

Joint activities

We want our PhD students to be a tight-knit group who work together towards a common goal. You will spend half your time in "RTG offices" near each other, and half your time in your primary advisor's group. You will also meet weekly for brainstorming sessions and reading groups.

Once a year, the whole RTG (students, PIs, and guests) will go on a one-week retreat, e.g. to Schloss Dagstuhl in the Northern Saarland. The program will be designed by the students to fit your interests and could e.g. involve a hackathon and joint research projects. You can also help decide what international guests to invite.

Coursework

We want our PhD students to talk to each other at a technical level, and therefore expect that every student has taken classes in at least two research areas.  If you have focused on a single area so far, you can learn about another one in our MSc-level courses.

Every student in the RTG will also take our award-winning course "Ethics for Nerds". AI increasingly affects our lives at many levels, and as researchers we must live up to our responsibility to having our technology used for good.

People skills

We want to prepare our students for a successful career after the PhD. Each student will typically take two courses per year (one or two days each) on topics such as scientific writing, presenting, self-management, or talking to the media. Students get to select their course program themselves, with support from the RTG office.

You will also have the opportunity to teach classes for advanced students, supervise students assistants or MSc theses, and participate in outreach activities. Of course teaching activities are totally voluntary.

Frequently Asked Questions

The RTG will fund three cohorts of PhD students in its first phase. A first group of six students will start in October 2023; a second cohort of six students in September 2024; and a third cohort of twelve students in September 2025.

In addition, we will associate 10-20 PhD students and postdocs working on neuroexplicit models in other research projects into the RTG. These students and postdocs will participate in RTG activities as well and contribute to the lively research environment we aim to create.

Most PhD students will be employed by Saarland University and receive four-year contracts at the E13 100% payscale. This translates to a yearly salary of roughly € 52.000. Conditions may differ slightly for students affiliated with an MPI or CISPA.

When you start your PhD, you will need to hold an MSc degree in computer science or a related discipline. You will also need to demonstrate mastery of English at the C1 level.

Notice that you do not need to hold the MSc degree when you submit your application; but you have to argue convincingly that your MSc will be finished when you join the RTG.

Students who joined the Saarbrücken Graduate School of Computer Science with a BSc degree are eligible for joining the RTG if they have successfully passed their Qualifying Exam.

Please submit your application by email to apply@neuroexplicit.org. Be sure to include the reference number mentioned in the job ad.

Please attach a single PDF file containing the following information:

  • A CV that describes your education and work history, any particular achievements, and your publications (if any).
  • Transcripts of your academic degrees and, if possible, a copy of your MSc and/or PhD diploma.
  • A research proposal (1-2 pages) that outlines your research interests with respect to the topic of the RTG.
  • Tell us what research area(s) interest you most (Language, Vision, Action, Foundations) and which up to three PIs you would most like to work with.
  • The names and email addresses of two academic references.

Note the privacy notice to see details about how we process your application data.

Once you submit your application, we will ask your academic references for letters of reference and review your application in detail.

We will then invite a shortlist of applicants to an in-person interview day in Saarbrücken. You will give a talk about your research so far and your research interests, and you will meet with some PIs one-on-one. We will reimburse your travel costs to Saarbrücken. If an in-person interview is not possible for you, we can also interview you over Zoom.

We are asking you to submit a research proposal as part of your application. This should be a brief (1-2 pages) statement of your research interests, so we can get a clearer idea of what you’d like to do in the RTG.

Feel free to include a brief summary of relevant previous work you have done, but the focus should be on ideas for future work. We are interested in what kind of research question you find attractive, and whether you are capable of motivating and articulating a research question. We want to see if you can substantiate your research idea by spelling out a concrete research project towards your overarching vision and some of the methods you would apply.

At the same time, please keep in mind that we are not asking you to write an ERC proposal: It is okay if your ideas are not fully fleshed out. Your actual PhD topic will emerge within your first six months in the RTG, in collaboration with your advisors; what you wrote in the research proposal may serve as a starting point, or it could end up being something totally different.

If you need some inspiration, you could check out the list of example topics in the research areas above.

Every student in the RTG will have two PIs as their PhD advisors. You will choose your primary advisor before joining the RTG, or at the latest within your first month after arriving. Often, a match between student and primary advisor will already emerge during the interview process. We aim to distribute students somewhat evenly across advisors, so consider a more junior advisor if your first-choice advisor already has lots of students.

Within your first six months in the RTG, you and your primary advisor will then work out your thesis topic and recruit a secondary advisor who is a good fit for this topic. The secondary advisor may be from the same research field as the primary advisor, or they could be from a different field for interdisciplinary thesis topics.

Your position as a PhD student will be funded for four years, and this cannot be easily extended. It is primarily your own responsibility to finish your research and write your thesis before the four years are up. Your advisors will support you in this.

However, we have built a few guardrails into the RTG that will make it more likely that you will actually finish within four years. After six months, you will write a thesis proposal and present it to the rest of the RTG. You will then meet once a year with your advisors and a third PI to discuss your achievements of the past year and your plans for the next year. This is an opportunity to regularly update your research agenda and identify and resolve any potential problems with your progress.

We believe strongly that PhD students benefit from spending time in different research environments, both in terms of broadening their scientific horizons and in terms of personal growth. We will therefore fund one three-month research visit by each PhD student to an international research site of their choice. We are furthermore supportive of students who want to spend time in a paid internship, and will extend the student’s contract by the duration of the internship.

The RTG will frequently host guests from all over the world. Students will benefit from spending time with these guests, and will be encouraged to propose guests who should be invited. The students will also get the opportunity to work with our Mercator Fellow, Ellie Pavlick, during her visits to Saarbrücken.

You will have access to ample compute resources and the opportunity to hire BSc and MSc students who will support your work as research assistants. If you like, you can teach a seminar about your research and supervise BSc or MSc theses, but this is totally voluntary.

In addition to the two PhD advisors, each student will also be assigned a PI as mentor. The mentor will be a point of contact on such topics as self-organization and career advice, and can help in case of conflicts. We furthermore aim to increase the proportion of female computer scientists, and will implement a number of measures to support the careers of our female PhD students.

Yes! The working language in the RTG will be English, and we anticipate that our PhD students will be a very international crowd. Even among our MSc students, about two thirds come from countries other than Germany.

You can survive in Saarbrücken as a speaker of English, but we recommend that you learn German while you are here to fully participate in the social and cultural life around you. The university offers language courses that you can attend.

Saarbrücken is the capital of the German state of Saarland and a decent-sized city of about 200.000 people. The French border is only a few kilometers away and within easy reach of the local light rail, and Luxembourg is an hour’s drive. You can get to Paris or Frankfurt in less than two hours on the train.

Saarbrücken has a very lively culture scene for a city of its size, ranging from opera and ballet to swing dancing and from the night life in the Nauwieser Viertel to chilling by the Saar river in the summer and going for hikes in the countryside. Saarbrücken hosts the annual Saar Music Festival and the Max Ophüls Film Festival, one of the most prestigious German-language movie festivals.

The living expenses in Saarbrücken are moderate, especially compared to bigger cities like Munich or Berlin, and you will live quite comfortably with your salary as a PhD student.