HAL Paris Dauphine-PSL
Not a member yet
21932 research outputs found
Sort by
RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning
No full textInternational audienceLarge language models (LLMs) deployed as agents solve user-specified tasks over multiple steps while keeping the required manual engagement to a minimum. Crucially, such LLMs need to ground their generations in any feedback obtained to reliably achieve the desired outcomes. We propose an end-to-end reinforcement learning method for teaching models to leverage execution feedback in the realm of code synthesis, where state-of-the-art LLMs struggle to improve code iteratively compared to independent sampling. We benchmark on competitive programming tasks, where we achieve new state-of-the art results with both small (8B parameters) and large (70B) models while reducing the amount of samples required by an order of magnitude. Our analysis of inference-time behavior demonstrates that our method produces LLMs that effectively leverage automatic feedback over multiple steps
Un dilemme de bien commun amusant fait-il apprendre ?
No full textUn jeu de rôles éducatif a été construit dans le cadre d’un projet OHM Bassin Minier de Provence, PollutionSolutions. Le projet « Autour du jeu » visait à tester ce jeu en contexte car il sera proposé en accès libre aux enseignants du secondaire dans le futur. Il donc était important de vérifier son adéquation aux programmes et sa capacité à générer les discussions anticipées par les concepteurs. Ce jeu est centré sur le thème de la pollution et utilise les ressorts comportementaux du dilemme de biens communs de la théorie des jeux en économie. Le projet s'inscrit dans la démarche recommandée par Elinor Ostrom qui est d'éduquer les jeunes générations à la compréhension de la tension entre intérêt collectif et intérêt individuel. Celle-ci, mise en scène dans le jeu, est expérimentée par les participants qui doivent la résoudre en établissant un dialogue. Le jeu a été testé dans différentes classes et dans plusieurs contextes : avec des préparations pédagogiques diverses et des groupes plus ou moins préparés en fonction des connaissances du caractère des élèves. Il a aussi été encore corrigé grâce à des ateliers avec des collègues spécialistes des jeux sérieux. Le processus a permis de mettre en avant un certain nombre de comportements émergents et de pistes de réflexion collective à mettre en place lors du debriefing. Un guide d'utilisation et des séquences pédagogiques ont été produites
Vanishing viscosity limit for hyperbolic system of conservation laws in 1-d with nonlinear viscosity: Part-I Uniform BV estimates
No full textWe consider the following parabolic approximation for hyperbolic system of conservation laws in 1-D with non-singular viscosity matrix and strictly hyperbolic, We prove global in time uniform bound for solution to this parabolic system when \varepsilon>0 provided that the initial data is small in and the matrix and commutate
Matrices totalement équimodulaires: décomposition et triangulation.
No full textTotally equimodular matrices generalize totally unimodular matrices and arise in the context of box-total dual integral polyhedra. This work further explores the parallels between these two classes and introduces foundational building blocks for constructing totally equimodular matrices. Consequently, we present a decomposition theorem for totally equimodular matrices of full row rank. Building on this decomposition theorem, we prove that simplicial cones whose generators form the rows of a totally equimodular matrix sa\-tisfy strong integrality decomposition properties. More precisely, we provide the Hilbert basis for these cones and construct regular unimodular Hilbert triangulations in most cases. We conjecture that cases not covered here do not exist
ELI Guidance on Company Capital and Financial Accounting for Corporate Sustainability
No full textInternational audienceThe European Law Institute – ELI Guidance proposes a set of Recommendations on company capital and financial accounting for corporate sustainability aimed at: (i) providing a frame of reference and analysis to understand corporate sustainability in the context of business and law; (ii) pointing to specific issues which need to be addressed by European and national lawmakers and regulators; and (iii) establishing a set of company law instruments which set out possible solutions to cope with these issues. The Recommendations aim at restating and modernising well-established principles of European company law on: (i) distributions; (ii) equity capital maintenance; and (iii) non-distributable reserves. Specific attention was paid to the enhanced controlling of new kinds of distributions such as share buybacks, as well as to limiting distributions of non-realised gains. The Project Team developed a comprehensive set of Recommendations aimed at fostering and facilitating sustainable business conduct through responsible company capital management and financial accounting adjustments. Further Recommendations were provided on related policy and regulatory matters concerning the EU framework for corporate sustainability. As a whole, the Recommendations propose that companies commit to a prudent use of resources, by setting aside sufficient reserves to meet social and environmental commitments over long-term horizons, and establish a fair balance between these commitments and distributions to shareholding investors. Corporate sustainability may be enhanced by implementing controls over distributions while reinforcing reserve provisioning. This in turn will ensure company continuity and resilience, as well as financial stability and sustainable development for the benefit of business and society at large
Probabilistic Prediction of Arrivals and Hospitalizations in Emergency Departments in Île-de-France
No full textInternational audienceForecasts of future demand is foundational for effective resource allocation in emergency departments (EDs). As ED demand is inherently variable, it is important for forecasts to characterize the range of possible future demand. However, extant research focuses primarily on producing point forecasts using a wide variety of prediction algorithms. In this study, our objective is to generate point and interval predictions that accurately characterize the variability in ED demand using ensemble methods that combine predictions from multiple base algorithms based on their empirical performance.MethodsData consisted in daily arrivals and subsequent hospitalizations at 72 emergency departments in Île-de-France from 2014-2018. Additional explanatory variables were collected including public and school holidays, meteorological variables, and public health trends. One-day ahead point and 80% interval predictions of arrivals and hospitalizations were produced by predicting the 10%, 50%, and 90% quantiles of the forecast distribution. Quantile prediction algorithms included methods such as ARIMAX, variations of random forests, and generalized additive models. Ensemble predictions were then formed using Exponentially Weighted Averaging, Bernstein Online Aggregation, and Super Learning. Prediction intervals were post-processed using Adaptive Conformal Inference techniques. Point predictions were evaluated by their Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE), and 80% interval predictions by their empirical coverage and mean interval width.ResultsFor point forecasts, ensemble methods achieved lower average MAE and MAPE than any of the base algorithms. All of the base algorithms and ensemble methods yielded prediction intervals with near optimal empirical coverage after conformalization. For hospitalizations, the shortest mean interval widths were achieved by the ensemble methods.ConclusionsEnsemble methods yield joint point and prediction intervals that adapt to individual EDs and achieve better performance than individual algorithms. Conformal inference techniques improves the performance of the prediction intervals
Competing for Being the Representative Field-Level Organization: When the Representative Role of A Meta-Organization is Contested by an Individual-Based Organization
No full textInternational audienceHow meta-organizations (MOs) can be engaged in competitive settings remains an underexplored issue, largely because scholars have traditionally emphasized MOs’ tendency toward monopoly (Ahrne & Brunsson, 2005) and focused on potential internal tensions between MOs and their members. However, it is not uncommon for an MO to find itself in competition with other organizations, including nonmember, individual-based organizations. In this paper, adopting an MO theory perspective and drawing on insights from the literatures on competition and representational legitimacy, we investigate competitive tensions within a health policy-related field. Our research question is as follows: how does representational legitimacy become a central object of competition when the representative role of an MO is contested by another organization in a public-policy-related field? Our empirical study focuses on two organizations – one an MO, the other an individual-based organization – that compete for status and authority, ultimately seeking recognition by public authorities as the central, if not official, representative of their field. We highlight the importance of representational legitimacy alongside more classical dimensions of authority based on expertise and knowledge. We also emphasize the meta-organizational form as a distinct type of representative structure, owing to its specific membership composition. Finally, we outline the central role played by policymakers as pivotal third and fourth actors in this competition, having created the conditions for its emergence and persistence
Les leçons marketing d’un chef d’œuvre télévisuel : Mad Men
No full textDepuis la rentrée 2025, ARTE propose gratuitement l’intégrale de la série Mad Men. L’occasion idéale de (re)voir une série acclamée par la critique et d’en apprendre beaucoup sur l’histoire du marketing, ses évolutions et ses dérives