This report covers the work performed during the first reporting period (January 2024 – December 2025) of the DeMythif.AI COFUND programme, an international doctoral training programme in Artificial Intelligence led by Université Paris-Saclay through the Institut DataIA, with 19 academic and industrial partners. The programme focuses on three scientific pillars: uncertainty quantification, AI explainability, and frugality (data and energy efficiency). It describes the two selection processes followed to recruit PhD fellows across two calls, the training and research environment set up for the fellows, and any deviations from the Grant Agreement.
The recruitment process for Call 1 started in September 2023 with an internal call for PhD projects open to academic and industrial partners of the programme, and ended in May 2024 with the selection of fellows after a rigorous two-stage evaluation (written pre-selection followed by interviews). 40 PhD projects were submitted and 500 applications from 200 unique candidates were received. The pre-selection and selection committees, composed of 14 external and international experts chaired by Cécile Germain, evaluated candidates on academic excellence (60%), extra-professional skills (20%), and motivation and career plans (20%). The overall quality of the applications was high, reflecting the attractiveness of the programme and of Université Paris-Saclay's research ecosystem. 8 fellows ultimately started their PhDs in October 2024, covering Computer Science, Mathematics, Physics and Engineering. Call 2 followed the same methodology with improvements based on lessons learned, offering 21 positions across 48 PhD topics. 300 applications were received, 62 candidates were interviewed, and 18 fellows were enrolled starting October 2025.
A total of 26 PhD fellows from 18 nationalities across all continents were recruited in two cohorts. An important communication effort was made through Euraxess, ADUM, the Association Bernard Grégory, LinkedIn, institutional websites, Graduate School networks, email campaigns to international master's programmes, alumni networks, and the academic networks of researchers at Université Paris-Saclay. See reports D3.1 D3.2 and D3.8. The recruited fellows were immediately immersed in a programme-specific environment: Welcome Days were organised for each cohort (November 2024 and November 2025), including dedicated courses on Open Science, Data Management Plan writing with personalised coaching, and the responsible use of GenAI in scientific research. Fellows have access to a wide catalogue of PhD-level AI courses (9 modules, approximately 20 hours each) managed by the Graduate School "Engineering and Systems Sciences", mandatory training in scientific ethics and discrimination prevention, the Academic Writing Center, French as a Foreign Language courses, and career development pathways. Secondment discussions have started for the first cohort with industrial partners (EDF, GE Healthcare, SLB) and international collaborations (CERN, ETS Montréal, Google). See reports D5.1 and D5.2 for ethics compliance. The first scientific output emerged in October 2025, with Raffael Schön having a paper accepted at a workshop of the International Conference on Computer Vision (ICCV) 2025.
Several deviations from the Grant Agreement occurred during this period. Call 1 recruited 8 fellows instead of the planned 15, due to French national security clearance restrictions (ZRR), competition from candidates' home countries, and the unilateral withdrawal of a candidate by a host laboratory that had secured full alternative funding. The corrective action was to increase recruitment targets for Call 2, which successfully recruited 18 fellows (vs. 15 initially planned), partially compensating for the shortfall. Industry partner involvement was lower than expected, as the COFUND co-funding model was perceived as less attractive than the French CIFRE scheme; corrective actions include stronger integration of industry partners in secondments, training and events. The hiring of a dedicated project manager was significantly delayed by the severe cyberattack on Université Paris-Saclay in August 2024 and by successive recruitment setbacks, but the position was ultimately filled in June 2025. Despite these challenges, the programme successfully enrolled both cohorts and established a fully operational training and research environment for its 26 fellows.
