Amid today’s environmental challenges, the long-standing Western divide between nature and culture is increasingly being called into question. Roman Antiquity, a significant period in our past, helps us reconsider such opposition. For the Romans, maintaining a harmonious relationship with ‘nature’ was essential to sustaining order and prosperity, as it was not a mere physical environment, but a subversive force, a realm intertwining the sensory with the divine. The use of real and iconographic vegetation in the living spaces was a central concern. The VegArch project aims to explore this vegetalisation within the specific context of Gallo-Roman sanctuaries of the 1st-3rd c. AD. From the penetration of Italic traditions to an unprecedented ‘hypervegetalisation’ of architecture, this greening has long been explained by a presumed persistence of Celtic naturalism or a mere decorative trend. Its role in shaping the experience of nature, the cosmos, and the divine can now be addressed with the expanding archaeological data and ecofacts and through an interdisciplinary approach. The vegetal forms will be gathered and compared in a trans-medial analysis, considering real and artificial forms as one and the same vegetal art and the sanctuary as a unitary system. This formal step will highlight the specificities of Gallo-Roman contexts in a global/local perspective. Through comprehensive and contextualised case studies, the vegetal forms will then be analysed in their spatial, temporal and sensory dimensions to reveal their polysemic value and how they engage the senses and shape the experience of the sacred. The resulting synthesis will contribute to the cultural history of vegetalisation of Western architecture, helping to unveil the relationship with nature in ancient societies, questioning our own practices and beliefs. The VegArch project, for which ENS is the ideal institution, will open numerous new perspectives and foster long-term European collaborations.

Learning and adaptation play a central role in interacting with a changing environment, yet the neural substrate underlying such flexible, goal-directed behaviors has remained elusive. Neuroscience experiments have traditionally focused on how individual brain regions perform abstract, highly simplified tasks. However, recent technological advances, coupled with powerful AI-accelerated software, have rapidly enabled the monitoring of large populations of neurons over many days, across multiple brain regions, and during increasingly complex, naturalistic behaviors. Yet even with such data within our reach, we still lack the theoretical and quantitative tools that are necessary to understand the fundamental principles guiding learning in neural populations. DULCE aims to fill this gap by establishing a unified framework to understand learning in complex environments. The core hypothesis of DULCE is that in naturalistic conditions, learning engages multiple co-occurring learning processes that are distributed across the brain, and which work together to reshape neural dynamics to perform new tasks. As such, DULCE aims to uncover the behavioral, population- level, and synaptic learning rules responsible for guiding learning in complex environments. By interweaving statistical modelling, dynamical systems theory and machine learning, DULCE will: i) Develop hierarchical models of behavior that can disentangle the rules governing simultaneously occurring learning processes. ii) Provide a unified theory of how region-specific learning rules in the cortex, cerebellum, and striatum coordinate to form a distributed learning system. iii) Develop interpretable dimensionality reduction methods to identify the rules governing how task-relevant dynamics evolve in large-scale neural data over learning. Through this three-pronged attack, DULCE aims to lay the foundation necessary to uncover the neural mechanisms controlling the Dynamics Underlying Learning in Complex Environments.

The spatial arrangement of solids and pore spaces within soil (soil structure) is a major source of variability in the hydraulic properties of soils. It is through changes in soil structure that biological activity, land management, or the passage of time (including wet-dry and freeze-thaw cycles) affect the capacity of soil to retain and conduct moisture. Nevertheless, soil structure is at most tentatively represented in Earth-system models. As a result, we cannot quantify the effect of biological activity and management on soil water content and thus their full effect on the regional climate or the land carbon sink. This represents an important gap in our understanding of the natural environment, as well as a potential blind spot in our response to the climate emergency. Including soil structure in land-surface models is difficult, because its effect on the soil hydraulic properties cannot be reliably predicted based on a single, easy-to-measure quantity such as bulk density. At the same time, more detailed soil structure data and the methods to incorporate it are largely missing. MOSS will address these issues by 1)~developing a physics-based method of incorporating beyond-density soil structure information into the models of soil hydraulic properties, 2)~implementing this method in a state-of-the-art Earth-system model, and 3)~using the new model to quantify the impact of soil structure on the land carbon sink. By doing so, MOSS will likely demonstrate that soil structure must be included in numerical models and protected in the field. Moreover, the method developed by MOSS will be applicable to agronomical and hydrological models and thus will have a plausible impact on the socially and economically important policy decisions these models inform. Finally, because of the interdisciplinary background of the researcher, we expect that the high visibility and copious training program offered by the fellowship will have a defining impact on their scientific career.

Liquid-liquid phase separation (LLPS) is central to compartmentalisation of biochemical processes and allows co-localisation of a whole biological machine and its substrates at high local concentrations. This often dynamic and reversible assembly is formed by multivalent interactions between several biomolecules, and some instances involve low complexity sequences that have been linked to amyloid fibre formation. While the biophysical understanding of this phenomenon has recently been of high interest in the scientific community, the interactions of LLPS-forming proteins from the dilute phase with the interface to the condensed phase remain elusive. In this project we aim to dissect these transient interactions using state-of-the-art biophysical techniques. More specifically, we will use nuclear magnetic resonance (NMR), high-resolution-relaxometry (HRR) and an array of single-molecule fluorescence techniques to dissect up to atomic resolution and at multiple time-scales the transient interactions of the dilute phase proteins with the interface of the condensed state. We shall rely on the non-homologous end joining (NHEJ) system, that our laboratory has recently shown to exhibit LLPS in a broad range of conditions. Atomic-level dynamic information on the mechanisms for NHEJ phase separation and assembly could prove crucial both in the fundamental understanding of LLPS formation and growth, and in rational drug design aimed at preventing double-strand break repair by NHEJ in the frame of cancer treatment.

The proposed project will focus on the well-posedness theory for hyperbolic systems of conservation laws in multiple space dimensions. The project will consider the stability and well-posedness theory for such systems, in the cases with and without source. In his 2023 survey of the field, Dafermos writes that "in regard to systems [of conservation laws], in one spatial dimension, the fundamental question whether the Cauchy problem in the BV setting is well-posed for initial data with large total variation remains wide open.'' Our program to study quantitative stability will allow us to consider not only large BV solutions, but also large L^2 solutions with "infinite BV." Classically, the best theory of well-posedness for hyperbolic systems in one space dimension is the L^1 theory of Bressan and coworkers, which considers small-BV solutions. Recent results of the researcher, Chen, and Vasseur go significantly beyond the classical small-BV theory, and are able to treat even large L^2 data. The theory uses the technique of a-contraction, and the key assumption is a strong trace condition, a regularity assumption strictly weaker than BV_loc. Our first main objective in this proposal is to quantify the stability in the a-contraction theory, which hasn't been done so far. More precisely, Objective 1 (Quantitative stability): In the setting of large data, derive quantitative stability estimates between L^2 and BV solutions for a large class of systems in one space dimension. The strong trace condition is the key boundary between general weak solutions and the solutions we can show uniqueness for. This brings us our Objective 2 (Regularity): For scalar conservation laws, possibly with nonlocal or unbounded source, under mild technical assumptions, show the existence of the strong traces. This is open even in the one dimensional scalar case with unbounded source. Showing the existence of strong traces would be a significant step towards the program of Dafermos.