Extended Reality (XR) experiences involve interactions between users, the real world, and virtual content. A key step to enable these experiences is the XR headset sensing and estimating the user's pose in order to accurately place and render virtual content in the real world. XR headsets use multiple sensors (e.g., cameras, inertial measurement unit) to perform pose estimation and improve its robustness, but this provides an attack surface for adversaries to interfere with the pose estimation process. In this paper, we create and study the effects of acoustic attacks that create false signals in the inertial measurement unit (IMU) on XR headsets, leading to adverse downstream effects on XR applications. We generate resonant acoustic signals on a HoloLens 2 and measure the resulting perturbations in the IMU readings, and also demonstrate both finegrained and coarse attacks on the ORB-SLAM3 and an open-source XR system (ILLIXR). With the knowledge gleaned from attacking these open-source frameworks, we demonstrate four end-to-end proof-of-concept attacks on a HoloLens 2: manipulating user input, clickjacking, zone invasion, and denial of user interaction. Our experiments show that current commercial XR headsets are susceptible to acoustic attacks, raising concerns for their security.

Small language models (SLMs) struggle to learn complex reasoning behaviors, especially when high-quality traces are scarce or difficult to learn from. A typical approach for training such models combines a supervised fine-tuning (SFT) stage, often to distill reasoning capabilities from a larger model, followed by a reinforcement learning (RL) stage such as Group Relative Policy Optimization (GRPO). In this paper, we investigate the fundamental limitations of this SFT + RL paradigm and propose methods to overcome them. Using a toy student-expert model over Markov chains, we demonstrate that the SFT + RL strategy can fail completely when (1) the expert's traces are too difficult for the small model to express, or (2) the small model's initialization achieves exponentially sparse rewards as task complexity grows. To address these, we introduce BREAD, a GRPO variant that bridges SFT and RL via partial expert guidance and branch rollouts. When self-generated traces fail, BREAD adaptively inserts short expert prefixes/hints, allowing the small model to complete the rest of the reasoning path, and ensuring that each update includes at least one successful trace. This mechanism both densifies the reward signal and induces a natural learning curriculum. BREAD requires fewer than 40\% of ground-truth traces, consistently outperforming standard GRPO while speeding up the training by about 3$\times$. Importantly, we find that BREAD helps the model solve problems that are otherwise unsolvable by the SFT + RL strategy, highlighting how branch rollouts and expert guidance can aid SLM reasoning.

In world-locked mixed reality (MR), virtual content is locked in place with respect to the real world. Pose estimation is a key component to create world-locked MR experiences by estimating the device's position and orientation in order to render the virtual content accordingly. Current methods of evaluating world-locked MR include user studies, which are time consuming, and absolute trajectory error (ATE), which does not directly represent what is shown on the user's display. In this work, we propose ReplayAR, a tool that can replay user movement traces and output the corresponding visualizations (renderings) of the MR display. ReplayAR can be used to compare renderings from different MR pose estimation methods side by side, using our proposed Visual Difference metric. We implemented ReplayAR on a Hololens 2 MR headset and used it to evaluate open and closed-source pose estimation methods on standard datasets and our own collected traces. The results suggest that Visual Difference better reflects what is shown on the MR display compared to ATE. We hope that ReplayAR can encourage reproducible evaluation of world-locked MR, and towards this, we release the open-source code.

Recent successes in natural language processing have led to the proliferation of large language models (LLMs) by multiple providers. Each LLM offering has different inference accuracy, monetary cost, and latency, and their accuracy further depends on the exact wording of the question (i.e., the specific prompt). At the same time, users often have a limit on monetary budget and latency to answer all their questions, and they do not know which LLMs to choose for each question to meet their accuracy and long term budget requirements. To navigate this rich design space, we propose TREACLE (Thrifty Reasoning via Context-Aware LLM and Prompt Selection), a reinforcement learning policy that jointly selects the model and prompting scheme while respecting the user's monetary cost and latency constraints. TREACLE uses the problem context, including question text embeddings (reflecting the type or difficulty of a query) and the response history (reflecting the consistency of previous responses) to make smart decisions. Our evaluations on standard reasoning datasets (GSM8K, CSQA, and LLC) with various LLMs and prompts show that TREACLE enables cost savings of up to 85% compared to baselines, while maintaining high accuracy. Importantly, it provides the user with the ability to gracefully trade off accuracy for cost.