Andre Freitas
1 year ago
This position has expired. You can find similar positions from the section below or browse our positions listing pages.
Agent/LLM-based Evidence-based Reasoning IDIAP and EPFL PhD programme in Switzerland
Degree Level
PhD
Field of study
Computer Science
Funding
4-year position funded by the SNSF-FAPESP RATIONAL project
Deadline
December 31st
Country
Switzerland
University
University NameSuggested positions
University of St. Gallen
Switzerland
2 weeks ago
Postdoctoral Researcher position in Conversational AI and LLMs at the University of St.Gallen (School of Computer Science), Switzerland, within the SNSF-funded CREST-AI project. The project focuses on natural language processing , large language models , dialogue systems , natural language generation , and reinforcement learning . The successful candidate will help develop grounded, adaptive conversational systems that tailor explanations to users’ understanding, with legal language as a testbed...
2 weeks ago
University of St. Gallen
Switzerland
1 week ago
Postdoctoral positions in NLP, Human-Centered AI, and Conversational AI are available at the University of St. Gallen, Switzerland through the SNSF-funded interdisciplinary project CREST-AI . The project sits at the intersection of Natural Language Processing , Human-Centered AI , Conversational AI , Large Language Models , philosophy, and law. The research aims to develop interactive language technologies that can detect gaps and misconceptions in user understanding and provide adaptive, ground...
1 week ago
University of Basel
Switzerland
Posted Yesterday
The Rhine AI research group at the Department of Mathematics and Computer Science, University of Basel, invites applications for a fully funded PhD position in post-training diffusion language models. The project, funded by the Swiss National Science Foundation, is titled “Foundations of Post-Training for Reasoning in Diffusion Large Language Models.” The successful candidate will develop reinforcement learning methods for post-training large diffusion-based language models, investigate reasonin...
Posted Yesterday
University of St. Gallen
Switzerland
2 weeks ago
Postdoctoral Researcher position in NLP and Human-Centered AI at the University of St.Gallen (School of Computer Science), Switzerland, within the SNSF-funded CREST-AI project led by Prof. Christina Niklaus . The project focuses on user modeling , LLM reasoning , neuro-symbolic NLP , semantic representation, and interpretable models of users’ beliefs, uncertainty, and misconceptions from dialogue. A key goal is to develop human-centered evaluation methods for personalized AI and adaptive explana...
2 weeks ago
University of Basel
Switzerland
Posted Yesterday
The Rhine AI research group, led by Prof. I. Bogunovic in the Department of Mathematics and Computer Science at the University of Basel, Switzerland, invites applications for a postdoctoral position in machine learning and artificial intelligence. The project, funded by the Swiss National Science Foundation, focuses on the post-training of diffusion-based large language models (dLLMs) and improving their reasoning capabilities. The postdoctoral researcher will develop reinforcement-learning meth...
Posted Yesterday
Keywords
About this position
* How to make LLM-based models capable of rigorous, fact-based inference?
* How to build AI models which can make sense of large-scale complex evidence?
This is an exciting opportunity to work at the interface between Large Language Models (LLMs) and complex reasoning.
The project:
LLMs have defined the foundations for the construction of models which can interpret and reason over language at scale. However, technical challenges remain in delivering models which are capable of factually correct, controlled and rigorous reasoning, fundamental in critical domains of applications such as biomedicine and policy-making.
In this project, we will work on the development of novel natural language inference (NLI) paradigms which can jointly reason over qualitative and quantitative evidence at scale. The project will focus on areas such as: (i) the development of new reasoning planning models for evidence-based reasoning and (ii) the integration of NLI models with statistical, causal and mechanistic inference paradigms.
The candidate will work at the Neuro-symbolic AI Group at IDIAP and will also be affiliated with the EPFL PhD programme.
This is a 4-year position funded by the SNSF-FAPESP RATIONAL project in collaboration with Daniel Pedronette (UNESP).
Candidates are expected to have:
A BSc/MSc degree in Computer Science or related areas.
Previous academic or industrial project experience in NLP (evidenced by project results and papers).
Be confident in software development and in developing complex NLP experimental pipelines.
Interested applicants please send your CV to [email protected] by December 31st.
Funding details
4-year position funded by the SNSF-FAPESP RATIONAL project
ApplyKite TwinLearns as you applyMake your next move.
Prepare your application
Review drafts and confirm requirements with the university.
Professors
Researcher at Idiap Research Institute | Senior Lecturer (Associate Professor) at the University of Manchester | AI Group Leader at the digital Experimental Cancer Medicine Team
Switzerland
Research Interests
Computational Linguistics
Artificial Intelligence
Natural Language Processing
Nlp
Large Language Models
ApplyKite TwinLearns as you applyMake your next move.
Prepare your application
Review drafts and confirm requirements with the university.
I’ve participated in discussions about the limitations of Large Language Models (LLMs), with the most prominent critique being their alleged inability to reason.
LLMs possess clear spatially-abstracted representations of words, both in and out of context. This means they encode relationships between words and concepts in a multidimensional space. Language itself, however, is merely a medium for conveying information. What’s truly interesting is that when an LLM creates connections between abstract representations of concepts, facts, or thoughts, it is effectively processing and reasoning through those representations. This process aligns with the essence of reasoning: the ability to form connections. In fact, the word "intelligence" derives from the Greek intel-lego, meaning "to connect."
This perspective challenges the notion that LLMs lack reasoning. While their reasoning may differ fundamentally from human cognition, it’s difficult to deny the sophisticated abstraction and pattern recognition they exhibit.
On these topics, I recommend a PhD position in ???? shared by Andre Freitas , whose lectures I had the privilege of attending. His perspective adds depth to this debate and is worth exploring further, along with the Evidence-based nature of the LLM fact-controlled reasoning capabilities this PhD path is targeting, valuable for critical domains such as biomedicine and policy-making.