As generative artificial intelligence (genAI) tools become embedded in computing education workflows, it is essential to understand how students use such systems to learn beyond introductory programming. This work investigates the relationship between the use of genAI by students and their conceptual understanding of mathematical and algorithmic principles in an undergraduate machine learning course with 134 students. We deploy a course-specific, custom-interfaced large language model (LLM), CubBot, to examine (1) how students interact with genAI in an upper-level CS course via an analysis of anonymized chat logs and (2) how genAI usage relates to students' conceptual understanding and learning outcomes via a randomized, controlled assessment comparing performance with and without CubBot access. This research contributes to the growing body of work on genAI-supported education by providing one of the first empirical investigations into genAI's relationship with conceptual learning in an upper-level CS course.

In this session, we aim to foster a community of computer science faculty across many primarily undergraduate institutions. We will foster an inclusive environment through activities that allow for multiple forms of engagement (e.g. shows of hands, small group introductions, and full group discussion). We will facilitate discussions about ways computer science is changing at PUIs and continue to brainstorm ways that our community can be sustained outside of the SIGCSE TS.

Although pre-training on a large amount of data is beneficial for robot learning, current paradigms only perform large-scale pretraining for visual representations, whereas representations for other modalities are trained from scratch. In contrast to the abundance of visual data, it is unclear what relevant internet-scale data may be used for pretraining other modalities such as tactile sensing. Such pretraining becomes increasingly crucial in the low-data regimes common in robotics applications. In this paper, we address this gap by using contact microphones as an alternative tactile sensor. Our key insight is that contact microphones capture inherently audio-based information, allowing us to leverage large-scale audio-visual pretraining to obtain representations that boost the performance of robotic manipulation. To the best of our knowledge, our method is the first approach leveraging large-scale multisensory pre-training for robotic manipulation. For supplementary information including videos of real robot experiments, please see https://sites.google.com/view/hearing-touch.

As teaching-focused positions proliferate and university teaching careers become more professionalized, there is growing attention being paid to how teaching faculty are created. However, how teaching faculty are hired also deserves scrutiny.

In this session, we aim to foster a community of computer science faculty across many primarily undergraduate institutions. We will foster an inclusive environment through activities that allow for multiple forms of engagement (e.g. shows of hands, small group introductions, and full group discussion). We will facilitate discussions about ways computer science is changing at PUIs and brainstorm ways that our community can be sustained outside of the SIGCSE TS.

Three challenges limit the progress of robot learning research: robots are expensive (few labs can participate), everyone uses different robots (findings do not generalize across labs), and we lack internet-scale robotics data. We take on these challenges via a new benchmark: Train Offline, Test Online (TOTO). TOTO provides remote users with access to shared robots for evaluating methods on common tasks and an open-source dataset of these tasks for offline training. Its manipulation task suite requires challenging generalization to unseen objects, positions, and lighting. We present initial results on TOTO comparing five pretrained visual representations and four offline policy learning baselines, remotely contributed by five institutions. The real promise of TOTO, however, lies in the future: we release the benchmark for additional submissions from any user, enabling easy, direct comparison to several methods without the need to obtain hardware or collect data.

We report on a reformulated general Ethics and Robotics course, in which we aim to address the twin curricular challenges of exposing computer science students to ethics discourse and establishing a pathway for ethics-oriented modules to be designed into numerous computer science courses across an institution. Given computer science instructors' lack of time and expertise to build ethics modules themselves, we tasked our students with creating ethics modules for instructors of 11 computer science courses at our university. Our course participants represented a diverse range of backgrounds and perspectives that catalyzed lively discussions and creative ideas for ethics pedagogy innovation. We report on course details, including in-class activities, assignments, and the project. We discuss our findings, including reception from students and computer science instructors and planned updates for the next course iteration. Given the course's overall success, we share with the hope that others may learn from or adopt our course approach. Materials are available on our website: https://vdean.github.io/16-735-ethics-robotics.html.

Common approaches for task-agnostic exploration learn tabula-rasa --the agent assumes isolated environments and no prior knowledge or experience. However, in the real world, agents learn in many environments and always come with prior experiences as they explore new ones. Exploration is a lifelong process. In this paper, we propose a paradigm change in the formulation and evaluation of task-agnostic exploration. In this setup, the agent first learns to explore across many environments without any extrinsic goal in a task-agnostic manner.Later on, the agent effectively transfers the learned exploration policy to better explore new environments when solving tasks. In this context, we evaluate several baseline exploration strategies and present a simple yet effective approach to learning task-agnostic exploration policies. Our key idea is that there are two components of exploration: (1) an agent-centric component encouraging exploration of unseen parts of the environment based on an agent’s belief; (2) an environment-centric component encouraging exploration of inherently interesting objects. We show that our formulation is effective and provides the most consistent exploration across several training-testing environment pairs. We also introduce benchmarks and metrics for evaluating task-agnostic exploration strategies. The source code is available at https://github.com/sparisi/cbet/.

Exploration is one of the core challenges in reinforcement learning. A common formulation of curiosity-driven exploration uses the difference between the real future and the future predicted by a learned model. However, predicting the future is an inherently difficult task which can be ill-posed in the face of stochasticity. In this paper, we introduce an alternative form of curiosity that rewards novel associations between different senses. Our approach exploits multiple modalities to provide a stronger signal for more efficient exploration. Our method is inspired by the fact that, for humans, both sight and sound play a critical role in exploration. We present results on several Atari environments and Habitat (a photorealistic navigation simulator), showing the benefits of using an audio-visual association model for intrinsically guiding learning agents in the absence of external rewards. For videos and code, see https://vdean.github.io/audio-curiosity.html.