Enhancing the quality of experience (QoE) in interactive video streaming (IVS) remains a persistent challenge due to the need for ultra-low latency and rising bandwidth demands. Conventional algorithms, whether rule-based or learning-based, are obsessed with achieving tight coupling between encoding and sending bitrate adaptations for low-latency guarantee. However, our measurement studies reveal alarming harms of tight coupling in suppressing throughput, encoding bitrates and smoothness, as application- and transport-layer bitrate adaptations inherently have different mechanisms and goals. To tackle this problem, we propose Octopus, the first loosely coupled cross-layer bitrate adaptation algorithm for IVS to maximize QoE. Instead of blind synchronization, Octopus promotes mutual cooperation and independence between encoding and sending bitrate adaptations by integrating a multi-head network with shortcut connections and auto-regressive action modules. Additionally, based on meta-imitation reinforcement learning, we design a network condition-aware online adaptation scheme that enables the loosely coupled policy to swiftly adapt to diverse and dynamic wireless networks. We implement Octopus on a testbed, a microcosm of real-world deployment, with transceiver pairs running WebRTC on the WeChat for Business dataset. Results show that Octopus outperforms state-of-the-art algorithms, either improving bitrates by 37.1%, or optimizing stalling rate and smoothness by 54.1% and 9.2%, or achieving all-around improvements.

To tackle long-horizon tasks, recent hierarchical vision-language-action (VLAs) frameworks employ vision-language model (VLM)-based planners to decompose complex manipulation tasks into simpler sub-tasks that low-level visuomotor policies can handle. Typically, the VLM planner needs finetuning to learn to decompose a new task, which requires target task demonstrations segmented into sub-tasks by either human annotation or heuristic rules. However, without prior knowledge, the heuristic sub-tasks can deviate significantly from the visuomotor policy's training data, thereby degrading task performance. To address these issues, we propose a Retrieval-based Demonstration Decomposer (RDD) that automatically decomposes video demonstrations into sub-tasks with prior by aligning the visual features of the decomposed sub-task intervals with those from the training data of the low-level visuomotor policies. RDD outperforms the state-of-the-art sub-task decomposer on both simulation and real-world tasks, demonstrating robustness across diverse settings. Code and more results are available at https://rdd-neurips.github.io

The sheer volume of scientific experimental results and complex technical statements, often presented in tabular formats, presents a formidable barrier to individuals acquiring preferred information. The realms of scientific reasoning and content generation that adhere to user preferences encounter distinct challenges. In this work, we present a new task for generating fluent and logical descriptions that match user preferences over scientific tabular data, aiming to automate scientific document analysis. To facilitate research in this direction, we construct a new challenging dataset CTRLSciTab consisting of table-description pairs extracted from the scientific literature, with highlighted cells and corresponding domain-specific knowledge base. We evaluated popular pre-trained language models to establish a baseline and proposed a novel architecture outperforming competing approaches. The results showed that large models struggle to produce accurate content that aligns with user preferences. As the first of its kind, our work should motivate further research in scientific domains.1

Real-time semantic segmentation (SS) is a major task for various vision-based applications such as self-driving. Due to the limited computing resources and stringent performance requirements, streaming videos from camera-embedded mobile devices to edge servers for SS is a promising approach. While there are increasing efforts on task-oriented video compression, most SS-applicable algorithms apply more uniform compression, as the sensitive regions are less obvious and concentrated. Such processing results in low compression performance and significantly limits the capacity of edge servers supporting real-time SS. In this paper, we propose STAC, a novel task-oriented DNN-driven video compressive streaming algorithm tailed for SS, to strike accuracy-bitrate balance and adapt to time-varying bandwidth. It exploits DNN's gradients as sensitivity metrics for fine-grained spatial adaptive compression and includes a temporal adaptive scheme that integrates spatial adaptation with predictive coding. Furthermore, we design a new bandwidth-aware neural network, serving as a compatible configuration tuner to fit time-varying bandwidth and content. STAC is evaluated in a system with a commodity mobile device and an edge server with real-world network traces. Experiments show that STAC can save up to 63.7–75.2% of bandwidth or improve accuracy by 3.1–9.5% compared to state-of-the-art algorithms, while capable of adapting to time-varying bandwidth.

Maximizing quality of experience (QoE) for interactive video streaming has been a long-standing challenge, as its delay-sensitive nature makes it more vulnerable to bandwidth fluctuations. While reinforcement learning (RL) has demonstrated great potential, existing works are either limited by fixed models or require enormous data/time for online adaptation, which struggle to fit time-varying and diverse network states. Driven by these practical concerns, we perform large-scale measurements on WeChat for Business’s interactive video service to study real-world network fluctuations. Surprisingly, our analysis shows that, compared to time-varying network metrics, network sequences exhibit noticeable short-term continuity, sufficient for few-shot learning requirement. We thus propose Fiammetta, the first meta-RL-based bitrate adaptation algorithm for interactive video streaming. Building on the short-term continuity, Fiammetta accumulates learning experiences through offline meta-training and enables fast online adaptation to changing network states through few gradient updates. Moreover, Fiammetta innovatively incorporates a probing mechanism for real-time monitoring of network states, and proposes an adaptive meta-testing mechanism for seamless adaptation. We implement Fiammetta on a testbed whose end-to-end network follows the real-world WeChat for Business traces. The results show that Fiammetta outperforms prior algorithms significantly, improving video bitrate by 3.6%-16.2% without increasing stalling rate.

Offloading computing to edge servers is a promising solution to support growing video understanding applications at resource-constrained IoT devices. Recent efforts have been made to enhance the scalability of such systems by reducing inference costs on edge servers. However, existing research is not directly applicable to pixel-level vision tasks such as video semantic segmentation (VSS), partly due to the fluctuating VSS accuracy and segment bitrate caused by the dynamic video content. In response, we present Penance, a new edge inference cost reduction framework. By exploiting softmax outputs of VSS models and the prediction mechanism of H.264/AVC codecs, Penance optimizes model selection and compression settings to minimize the inference cost while meeting the required accuracy within the available bandwidth constraints. We implement Penance in a commercial IoT device with only CPUs. Experimental results show that Penance consumes a negligible 6.8% more computation resources than the optimal strategy while satisfying accuracy and bandwidth constraints with a low failure rate.