Shared service environments are typically designed to facilitate comfort, efficiency, and positive consumer experiences. In post-crisis contexts, however, these environments may also be appraised through concerns about exposure, controllability, and personal vulnerability. Building on a context-specific refinement of the organismic component within the Stimulus–Organism–Response framework, this research examines how residual infection-related fear is associated with consumer responses to shared service environments in a post-pandemic context.A two-study design was implemented within a single French-language survey (N = 425). Study 1 tested whether residual infection-related fear was associated with avoidance of public service environments through differentiated forms of health anxiety. Structural equation modeling showed that residual infection-related fear was positively associated with both personal and interpersonal health anxiety, but only personal health anxiety was associated with avoidance, yielding a significant indirect effect through the self-focused pathway. Study 2 experimentally manipulated store density (high vs. low) and sanitary signals (present vs. absent) to examine how servicescape cues influenced perceived crowding and shopping satisfaction. Results showed that higher density reduced satisfaction, but this negative effect was attenuated when sanitary signals were present. Perceived crowding did not mediate the density–satisfaction relationship. A separate exploratory analysis indicated that the overall association between sanitary signals and satisfaction was more positive among individuals reporting higher residual infection-related fear.Taken together, the findings suggest that post-crisis responses to shared service environments may reflect the interplay between situational cues and heterogeneous vulnerability orientations, consistent with a vulnerability-sensitive appraisal system. The research offers a context-specific contribution to servicescape theory by suggesting that, under residual health-threat salience, organismic processes may require a more differentiated specification in contexts marked by residual threat, and by highlighting how visible safety-related cues may improve evaluations of dense service environments. Exploratory evidence further suggests that the overall association between such cues and satisfaction is more positive among consumers who remain more sensitive to infection-related risk.
Transformer-based neural networks have emerged as powerful tools for combinatorial optimization problems, such as the Traveling Salesman Problem (TSP). However, their high computational demands during training raise concerns about scalability. This paper explores the use of POPMUSIC, a fast heuristic, to replace resource-intensive training with lightweight edge scoring. The study compares several sampling techniques —greedy search, beam search, and randomized selection (inspired by ant colony optimization)— both guided by POPMUSIC-generated edge frequencies and transformer outputs. Additionally, it evaluates how a transformer model, trained on uniform TSP instances of fixed size, generalizes to clustered and larger instances. The results demonstrate that randomized construction consistently outperforms beam search for both POPMUSIC and transformer outputs. While the pre-trained transformer generalizes well to larger and structurally different instances, traditional heuristics still surpass neural networks for large-scale TSPs. The findings highlight promising directions for hybrid methods that combine neural scoring with advanced heuristic selection strategies.
This study aimed to systematically identify, synthesize, and evaluate measurement properties of patient-reported outcome measures (PROMs) of eHL in adult populations. A systematic review was conducted, considering studies reporting the development or validation of eHL instruments for adult populations. Four databases and grey literature were searched from January 2000 to 2024, with additional website searches up to 2022. Quality assessment, data analysis and synthesis followed the COSMIN methodology and findings were reported according to PRISMA 2020 guidelines. The GRADE framework was used to assess evidence quality. A total of 8558 citations were identified. Seven instruments, 89 articles and 3 reports were included in this review. The HL19-DIGI, DHLI, TeHLI, eHLQ, eHLA, and Lisane demonstrated sufficient ratings for aspects of content validity, albeit with varying levels of evidence, ranging from very low to high. Five instruments showed sufficient ratings for structural validity and internal consistency, but evidence on their reliability was insufficient. No information on responsiveness was mentioned in articles. The HL19-DIGI, DHLI, eHEALS, and eHLQ were the most frequently investigated instruments. This review identified 17 eHL instruments, of which seven demonstrated adequate content validity. However, insufficient evidence exists regarding psychometric properties for widespread implementation. It is strongly recommended that the content of these instruments be updated to reflect patients’ evolving use of eHealth services, and that further psychometrics evaluations be conducted systematically. PROSPERO CRD42021232765.
Neural machine translation (NMT) in the legal domain is a linguistically and conceptually demanding task, primarily due to the complexity of legal language and the high level of precision it requires. The recent emergence of reasoning-capable language models opens new possibilities for tackling such challenges. They add to a set of other previously proposed techniques to enhance the translation quality, which includes supervised fine-tuning and reinforcement learning. In this work, we perform a comparison between these various approaches. More particularly, we evaluate small language models such as Qwen3.5 4B, Qwen3.5 9B, and Gemma 3 12B enhanced with various re-training paradigms and compare their performances against frontier reasoning models. We focus on the Swiss legal system, which – with its unique multilingual statutes – offers a particularly challenging testbed for reasoning-augmented models. Our results show that the quality of small “base” models can be greatly enhanced, and that reinforcement learning with verifiable rewards can be applied to NMT in the legal domain and surpasses the translation quality of supervised fine-tuning. The performance of enhanced small models is close to the one of state-of-the-art reasoning models yet remains inferior. We also note that re-training paradigms yield diminishing returns as model size increase. The code and models are publicly available at https://github.com/aixiuxiuxiu/Legal-MT-SFT-RL.
This paper assesses the ability of large language models (LLMs) to translate texts that include inter-sentential dependencies. We use the English-French DiscEvalMT benchmark (Bawden et al., 2018) with pairs of sentences containing translation challenges for pronominal anaphora and lexical cohesion. We evaluate 12 LLMs from the DeepSeek-R1, GPT, Llama, Mistral and Phi families on two tasks: (1) distinguish a correct translation from a wrong but plausible one; and (2) generate a correct translation. We compare prompts that encourage chain-of-thought reasoning with those that do not. The best models take advantage of reasoning and reach about 90
