PurposeThis study takes an institutional logics perspective to examine competing stakeholder logics in a new publicly funded French accelerator to foster the development of entrepreneurial firms. The study aims to understand how the accelerator staff managed the different logics that emerged during the mentoring programme.Design/methodology/approachOver a period of eighteen months, the authors used a process-based research design to follow two cohorts comprising 20 mentoring dyads, engaging in 40 narrative interviews with mentors, mentees and business accelerator officials, as well as conducting overt participant observation and archival data analysis.FindingsThis emerging mentoring programme encountered tensions emanating from two conflicting logics. Confronted with institutional pressures for managing the programme, the accelerator team first endorsed a performance-driven logic, reflected in potential economic outputs for mentee firms. They gradually acknowledged that mentoring practice can also benefit from a community logic built on trust, mutual respect and reflexive thinking.Originality/valueBy revealing how competing logics were managed in the accelerator, the study brings two contributions to the literature on entrepreneurial mentoring. Firstly, it shows that the governance mechanisms in place contributed to creating and resolving the tensions that emerged in the accelerator regarding performance versus community logics. Secondly, the study shows how the consensus gradually reached was contingent on the changing perception of the role of mentoring among the various stakeholders in the accelerator.
Natural language interfaces offer a compelling approach for music recommendation, enabling users to express complex preferences conversationally. While Large Language Models (LLMs) show promise in this direction, their scalability in recommender systems is limited by high costs and latency. Retrieval-based approaches using smaller language models mitigate these issues but often rely on single-modal item representations, overlook long-term user preferences, and require full model retraining, posing challenges for real-world deployment. In this paper, we present JAM (Just Ask for Music), a lightweight and intuitive framework for natural language music recommendation. JAM models user-query-item interactions as vector translations in a shared latent space, inspired by knowledge graph embedding methods like TransE. To capture the complexity of music and user intent, JAM aggregates multimodal item features via cross-attention and sparse mixture-of-experts. We also introduce JAMSessions, a new dataset of over 100k user-query-item triples with anonymized user/item embeddings, uniquely combining conversational queries and user long-term preferences. Our results show that JAM provides accurate recommendations, produces intuitive representations suitable for practical use cases, and can be easily integrated with existing music recommendation stacks.
The CMOS image sensors are more and more used for low flux imaging applications. While several strategies have been proposed for increasing the pixel conversion gain, this study proposes to compare some of them in a conventional imaging process, without any process adjustment. The goal is to analyze the efficiency of the various strategies in improving the conversion factor (CVF) and to discuss their impact on pixel performances, such as the nonuniformity. After the analysis of the drawbacks, recommendations are given for improved designs.
In this article, we prove a generic lower bound on the number of O-orientable supersingular curves over Fp2 , i.e curves that admit an embedding of the quadratic order O inside their endomorphism ring. Prior to this work, the only known effective lower-bound is restricted to small discriminants. Our main result targets the case of fundamental discriminants and we derive a generic bound using the expansion properties of the supersingular isogeny graphs. Our work is motivated by isogeny-based cryptography and the increasing number of protocols based on O-oriented curves. In particular, our lower bound provides a complexity estimate for the brute-force attack against the new O-uber isogeny problem introduced by De Feo, Delpech de Saint Guilhem, Fouotsa, Kutas, Leroux, Petit, Silva and Wesolowski in their recent article on the SETA encryption scheme.
