Coordinates: 44°24′44.29″N 79°39′40.08″W / 44.4123028°N 79.6611333°W / 44.4123028; -79.6611333Georgian College is a College of Applied Arts and Technology in Ontario, Canada. It has 13,000 full-time students, including 4,500 international students from 85 countries, across seven campuses, the largest being in Barrie.
Canadian postsecondary institutions experienced rapid growth in international student enrolment over the past decade, followed by a sudden decline resulting from recent federal immigration reforms. This shift reveals the sector's vulnerability to regulation and has created a climate of uncertainty for institutions that rely on international tuition and global recruitment. In this context of demographic expansion and regulatory pressure, the need for culturally responsive pedagogy becomes increasingly important. This chapter examines how faculty are adapting their teaching to support multilingual, multicultural classrooms, and how instructors often act as informal cultural mediators performing significant but largely invisible pedagogical labour without formal training or recognition. This work argues that the convergence of economic dependency, student investment, and policy change creates an ethical imperative for pedagogical innovation that supports diverse learners and strengthens inclusive learning environments.
Canadian colleges play an increasingly vital role in applied research, supporting regional innovation through partnerships with industry, community, and government. Backed by programs like the Tri-Agency College and Community Innovation (CCI) fund, colleges have expanded beyond their teaching mandates to tackle real-world challenges through practical, solution-focused research. However, many institutions still face barriers due to limited internal resources, inconsistent faculty engagement, and administrative structures designed primarily for education. This paper outlines seven key strategies identified by Canadian college research leaders to build and strengthen applied research capacity: (a) focus on core funding programs, (b) align with federal priorities while exploring other sources, (c) engage local industry and community, (d) integrate research with student programming, (e) collaborate with internal departments, (f) leverage marketing to promote impact, and (g) learn from established institutions and associations. These strategies offer a roadmap for colleges at any stage of research development to enhance sustainability, relevance, and long-term impact.
Building adaptable AI systems requires effective coordination of specialized capabilities across diverse tasks. Low-rank adaptation (LoRA) enables modular expertise, but existing routing approaches may require auxiliary data, additional training, or autoregressive decoding. We propose JevSoup, a training-free framework separating System One expert routing from System Two execution. Using only the input and expert descriptions, Jev selects two experts through structured probabilities. JevSoup retains the leading expert's update, projects the second onto the orthogonal complement of the first update's row space, and combines them with equal weights. Across 14 PorTAL tasks and three Qwen3 scales, JepSoup achieves absolute gains of up to 1.19% in task-macro and 1.21% in sample-micro accuracy over the strongest evaluated external baselines. Our code is available at https://github.com/Leowang980/JevSoup.
“Vibe coding”—where users describe applications in natural language and AI agents autonomously handle all implementation—has rapidly emerged as a mainstream development paradigm. Yet existing benchmarks evaluate mainly agent’s ability at code completion or fixing bug, not end-to-end application creation from a user’s perspective. We introduce ViBench, the first open-source benchmark for evaluating AI agents on realistic, end-to-end web application development. ViBench comprises tasks derived from production traces across 15 applications spanning different domains, covering both zero-to-one creation and feature extension—the two core workflows of vibe coding. Unlike prior benchmarks, tasks are specified entirely through user-facing requirements with no implementation constraints, and evaluation occurs at the abstraction level of an end user interacting with the application. We contribute an adaptive automatic evaluator that uses REPL-based browser automation to verify arbitrary implementations without assuming anything about code structure, achieving 99% step-level agreement with human experts across 1,082 test steps. Evaluating nine models—five closed-source and four open-weight—across 105 artifacts, we find that even the best models are far from reliable: Opus 4.6 and GPT-5.2 lead with only 46% and 42% Pass@1—fully correct artifacts on the first try—while no open-weight model exceeds 12% Pass@1. Feature extension on a stable reference codebase proves easier than zero-to-one creation, but errors compound when models extend their own code: seven of nine models degrade in the Vibe-on-Vibe setting, with complete failure rates rising sharply. Notably, open-weight models that perform competitively on coding benchmarks such as SWE-bench struggle to generalize to end-to-end vibe coding, exhibiting complete failure rates up to 40%—underscoring that end-to-end application development demands capabilities beyond what current coding benchmarks measure. As vibe coding matures into mainstream development practice, ViBench provides the open-source infrastructure needed to measure progress, compare models, and guide practitioners toward reliable AI-assisted development, beyond just a “Vibe Check”.
Psycholinguistic research and linguistic theory suggest that comprehending continuous speech involves segmenting it into sentences. Yet, direct neuroimaging evidence for this process remains scarce, partly because sentence boundaries coincide with acoustic, prosodic, lexical, and syntactic changes. Here, we used whole-brain fMRI to test whether sentence boundaries elicit a neural response beyond these accompanying speech properties during spontaneous speech comprehension. We modelled BOLD responses to sentence-final words alongside continuous regressors characterizing accompanying speech properties. This approach revealed sensitivity to continuously varying features of speech across multiple brain regions. Notably, the acoustic envelope predicted BOLD responses in bilateral primary auditory cortices. Responses associated with prosodic breaks and emphasis were found in right temporal cortex, syntactic bottom-up node count in the left superior temporal sulcus, and word concreteness in the parahippocampal gyri. Critically, sentence-final words were associated with an increased activation in right temporal regions, after accounting for acoustic, prosodic, lexical, semantic and syntactic properties. These findings provide evidence for a neural response to sentence boundaries consistent with active sentence segmentation. More generally, this study demonstrates the value of continuous-regressor modelling for investigating speech comprehension with fMRI. ### Competing Interest Statement The authors have declared no competing interest. Israel Science Foundation, 543/25