Stage Case-Based Reasoning on Concepts for Sensible Explainable AI- Saclay-H/F

CEA; Saclay; On-site; EUR 13,000 - 17,000; Full time

Job summary

CEA List, Saclay, propose un stage en Case-Based Reasoning sur des concepts pour une IA explicable et sensible. Vous travaillerez dans un environnement de recherche de pointe, avec accès à des ressources et à une équipe dédiée, pour concevoir et évaluer des solutions abductives et des procédures décisionnelles basées sur des concepts.

Le stage vise à prototyper des méthodes explicables, à tester des composants sur CaBRNet et à contribuer potentiellement à une publication scientifique, dans un

Qualifications

• Master étudiant en informatique et/ou ML.
• Connaissance du langage Python.
• Capacité à travailler en équipe.

Responsibilities

• Explorer et esquisser une procédure de décision.
• Implémenter un proof-of-concept des modèles Concept Bottleneck dans CaBRNet.
• Évaluer l’approche sur des métriques existantes et des méthodes d’évaluation humaine.

Skills

Python

Machine learning

Teamwork

Education

Master of Science in Computer Science / ML

Tools

PyTorch

Git/GitLab

CaBRNet

Job description

Stage Case-Based Reasoning on Concepts for Sensible Explainable AI- Saclay-H/F

Domaine

Contrat

Stage

Stage Case-Based Reasoning on Concepts for Sensible Explainable AI- Saclay-H/F

Case-based reasoning is a popular framework for explainable AI. It exposes the
network decision process by relating inference-time inputs with
prototypes.
Recent work proposed to rely on the formalism of Abductive Explanations as to
guarantee that an prototype-based explanation is both minimal and sufficient.
Assigning meaning to prototypes is however a time-consuming task, as the various models tend to learn prototypes that do not carry sufficiently high-level meaning.
A parallel line of work is that of Concept Learning. By injecting concepts
meaningful to human operators, Concept Bottleneck Models
provide a principled way to map feature maps to annotated concepts, providing an
easy to understand - and contest - decision procedure.
Recent work display a renewed interest in those approaches, by exposing counterfactuals generation capabilities.

As an intern at the CEA, you will have the opportunity to work in a world-renowned research environment. Our teams consist of passionate and dedicated experts, providing an environment conducive to learning and collaboration. You will have access to state-of-the-art equipment and top-tier research resources to carry out your assignments. The work performed may potentially lead to a scientific publication.

Context :

The main objective of the internship is to investigate, design and implement an
abductive explanations and
concept-based decision procedure.

Whatdo we expect from you ?

The broad internship goals are:

• familiarization with the state-of-the-art;
• sketching a decision procedure on paper;
• implement a proof-of-concept of concept bottleneck models in the CaBRNet library;
• evaluate the approach on existing metrics and various human-in-the-loop evaluation methods.

#Cea List

CaBRNet

Profile :

• master student or equivalent (2nd/3rd engineering school year) in computer
  science and/or machine learning;
• some knowledge of the Python programming language;
• ability to work in a team;

Preferred

• knowledge of the PyTorch framework;
• used to work with version control systems (git) and software forges (GitLab);
• background on explainable AI;

Site

Saclay

Saclay

Langues

Bac+5 - Master of Science

Oui

01/03/2027

Référence

2026-42007

The French Atomic Energy Commission (CEA) is a key
player in research, development, and innovation. Drawing on the widely
acknowledged expertise gained by its 16,000 staff spanned over 9 research
centers with a budget of 4.1 billion Euros, CEA actively participates in more
than 400 European collaborative projects with a large number of academic
(notably as a member of Paris-Saclay University)
and industrial partners. Within the CEA Technological Research Division, the
CEA-List institute addresses the challenges coming from smart digital systems.
Among other activities, CEA List's Software Safety and Security Laboratory (LSL)
research teams design and implement automated analysis in order to make software
systems more trustworthy, to exhaustively detect their vulnerabilities, to
guarantee conformity to their specifications, and to accelerate their
certification. The lab recently extended its activities on the topic of AI
trustworthiness and gave birth to a new research group: AISER (Artificial
Intelligence Safety, Explainability and Robustness).