In video-on-demand services, each video title is deployed as a bitrate ladder—a set of pre-encoded representations with increasing bitrate and quality. This paper introduces NILO, a Nested and Iterative Ladder Optimization method for designing bitrate ladders in video streaming services. NILO balances the tradeoffs between user Quality of Experience (QoE) and Content Delivery Network (CDN) efficiency, while meeting the operational needs of large-scale production systems. Our contributions include a multi-objective optimization framework that captures QoE and CDN efficiency during ladder construction, and a novel method that uniquely integrates standard optimization components to address this complex problem. We evaluate NILO at a large video streamer using trace-driven simulations and an A/B test in production with over one million users. Our results show that NILO achieves significant efficiency gains while maintaining QoE comparable to highly tuned production ladders. Specifically, NILO reduces storage by approximately 20% and streaming rate by about 2%, with options for greater efficiency gains at the cost of modest QoE degradation.

Autonomous cyber-physical systems liker obots and self-driving cars could greatly benefit from using formal methods toreason reliably about their control decisions.However, beforea problem can be solved it needs to be stated.This requires writing af ormal physics model of the cyber-physical system, which is a complex task that traditionally requires human expertise and becomes ab ottleneck.This paper experimentally studies whetherL arge Language Models (LLMs) can automate the formalization process.A2 0 problem benchmark suite is designed drawing from undergraduate levelp hysics kinematics problems.In each problem, the LLM is provided with an atural language description of the objects' motion and must produce am odel in differentialg ame logic (dGL).The model is (1) syntax checked and iteratively refined based on parser feedback, and( 2) semantically evaluated by checking whether symbolically executing the dGL formula recovers the solution to the original physics problem.As uccess rate of 70% (best over 5s amples) is achieved.We analyze failing cases, identifying directions forf uturei mprovement.This provides afi rst quantitative baseline forL LM-based autoformalization from natural language to ah ybrid games logic with continuous dynamics.

The emergence of the Spatial Web -- the Web where content is tied to real-world locations has the potential to improve and enable many applications such as augmented reality, navigation, robotics, and more. The Spatial Web is missing a key ingredient that is impeding its growth -- a spatial naming system to resolve real-world locations to names. Today's spatial naming systems are digital maps such as Google and Apple maps. These maps and the location-based services provided on top of these maps are primarily controlled by a few large corporations and mostly cover outdoor public spaces. Emerging classes of applications, such as persistent world-scale augmented reality, require detailed maps of both outdoor and indoor spaces. Existing centralized mapping infrastructures are proving insufficient for such applications because of the scale of cartography efforts required and the privacy of indoor map data.

World-scale augmented reality (AR) applications need a ubiquitous 6DoF localization backend to anchor content to the real world consistently across devices. Large organizations such as Google and Niantic are 3D scanning outdoor public spaces in order to build their own Visual Positioning Systems (VPS). These centralized VPS solutions fail to meet the needs of many future AR applications-they do not cover private indoor spaces because of privacy concerns, regulations, and the labor bottleneck of updating and maintaining 3D scans. In this paper, we present OpenFLAME, a federated VPS backend that allows independent organizations to 3D scan and maintain a separate VPS service for their own spaces. This enables access control of indoor 3D scans, distributed maintenance of the VPS backend, and encourages larger coverage. Sharding of VPS services introduces several unique challenges-coherency of localization results across spaces, quality control of VPS services, selection of the right VPS service for a location, and many others. We introduce the concept of federated image-based localization and provide reference solutions for managing and merging data across maps without sharing private data.

Many modern-day XR devices (e.g. mobile headsets, phones, etc.) lack the computing resources required to render complex 3D scenes in real-time. Typically, to render a high-resolution scene on a lightweight XR device, 3D designers arduously decimate and fine-tune the objects. As an alternative, remote rendering systems can utilize powerful nearby servers to stream rendering results to a client. While this is a promising solution, it can introduce a variety of latency and reliability issues, especially under variable network conditions. In this paper, we present a distributed rendering system that combines both remote rendering and on-device, “local” rendering to add robustness to network fluctuations and device workloads. To maximize user QoE, our approach dynamically swaps an object’s rendering medium, adjusting for client workload, low frame rates, and several perceptual characteristics. To model these characteristics, we perform a study under simulated conditions to measure how users perceive latency and complexity differences between objects in a scene. Using the results of the study, we then provide an algorithm for choosing the optimal object rendering medium, based on rendering complexity as well as network and latency models, ensuring that a target frame rate will be met. Finally, we evaluate this algorithm on a prototype implementation that can provide cross-platform split rendering using web technologies.

Analysts frequently require data from multiple sources for their tasks, but finding these sources is challenging in exabyte-scale data lakes. In this paper, we address this problem for our enterprise's data lake by using machine-learning to identify related data sources. Leveraging queries made to the data lake over a month, we build a relevance model that determines whether two columns across two data streams are related or not. We then use the model to find relations at scale across tens of millions of column-pairs and thereafter construct a data relationship graph in a scalable fashion, processing a data lake that has 4.5 Petabytes of data in approximately 80 minutes. Using manually labeled datasets as ground-truth, we show that our techniques show improvements of at least 23% when compared to state-of-the-art methods.