Granite-speech LLMs are compact and efficient speech language models specifically designed for English ASR1and automatic speech translation (AST). The models were trained by modality aligning granite-3.3-instruct to speech on publicly available open-source corpora. Comprehensive benchmarking on English ASR shows that they outperform several competitors’ models that were trained on orders of magnitude more proprietary data, and they keep pace on English-to-X AST for major European languages, Japanese, and Mandarin. The speech-specific components are: a conformer acoustic encoder using block attention and self-conditioning trained with connectionist temporal classification, a windowed query-transformer speech modality adapter used to do temporal downsampling of the acoustic embeddings and map them to the LLM text embedding space, and LoRA adapters to further fine-tune the text LLM. The models are freely available on HuggingFace2under a permissive Apache 2.0 license.1The latest models (revision 3.3.2) support multilingual ASR in English, French, German, Spanish and Portuguese and bidirectional speech translation to and from English. This paper covers the initial English-only release.2https://huggingface.co/ibm-granite/granite-speech-3.3-2b (and…-8b).

Computer Science educators widely agree that ethics is a vital and underdeveloped part of the CS curriculum. Attempts to increase ethics content within undergraduate CS programs have faced challenges integrating material into the current coursework. I present an ethical framework applicable to the core Computer Science activities of programming and software development, with potential for extension into other CS subfields. The application of this framework to a specific educational intervention is reserved for future work. In this paper, I focus on the framework itself and its theoretical justification. By shifting emphasis from hard ethical quandaries and advanced CS products to mundane challenges faced by a front line software developer, this framework may allow instructors to more easily and effectively integrate ethics material into introductory CS coursework. While instructors may continue to apply prepared scenarios, the framework de-emphasizes those in favor of scaffolding student coding practices that maximize the frequency of practice with ethical skills.

We want to facilitate an interactive gathering of faculty from a variety of disciplines and institutions - including primary, secondary, and tertiary schools - who are interested in exchanging ideas related to the learning and teaching of ethical concepts and skills in a CS or related curriculum, and evaluating the success of those efforts. We understand "curriculum" broadly to include both course and lab instruction and out-of-class experiences. Participants will leave with a set of open-access teaching resources, guidance for how best to use those resources at their own institutions, and ideas for evaluating any courses, curriculum, or programs they develop as a result.

As the COVID-19 pandemic spread throughout the United States in Spring 2020, universities around the country conducted the unprecedented feat of moving essentially their entire curricula online. This abrupt mid-semester shift in modality was the first of its kind at scale, but it will not be the last. Now that our experiences during the pandemic have demonstrated that an online transition is possible, we may see similar transitions proposed during smaller scale crises, especially as an alternative to school closures. We present an experience report focusing on computer science instructors at a large Midwestern university during the Spring 2020 online transition: the challenges they faced, as well as the interventions they implemented. As more data and analysis is generated from the first mass, unwilling modality shift, it is nevertheless important to focus on the visceral reactions and lessons learned from teachers and students in the moment as they responded to an unprecedented situation. We therefore also present our analysis of critical challenges and preliminary pedagogical recommendations for consideration during future rapid transition "involuntarily online" semesters.

In the space between the well documented shortage of access to computer science educational (CSed) opportunities, especially those that broaden participation, and the increasing consensus that computer science knowledge is a fundamental modern literacy lie an eclectic collection of non-traditional and informal CSed projects. An increasingly popular form of interest-based CSed product are board games, which are often justified as interventions due to their 'unplugged' nature and their supposedly natural appeal to diverse audiences of children. Through an exploration of the popular and academic discourses around the CSed board game genre, we found over 30 games which specifically aim to teach core CS content, often with a wide variety of rationales for utilizing a board game format. This poster will display the results of our preliminary work investigating how designers position cardboard as the exciting new alternative for teaching computer science education. Our examination of game makers' documentation and related research suggests that further work is required to determine how well the board game medium serves the various justifications presented by the game makers in our sample.

Debugging is difficult. When software fails in production, debugging is even harder, as failure reports usually provide only an incomplete picture of the failing execution. We present a system that answers control-flow queries posed by developers as formal languages, indicating whether the query expresses control flow that is possible or impossible for a given failure report. We consider three separate approaches that trade off precision, expressiveness for failure constraints, and scalability. We also introduce a new subclass of regular languages, the unreliable trace languages, which are particularly suited to answering control-flow queries in polynomial time. Our system answers queries remarkably efficiently when we encode failure constraints and user queries entirely as unreliable trace languages.