In this letter, we explore the communication-control co-design of discrete-time stochastic linear systems through reinforcement learning. Specifically, we examine a closed-loop system involving two sequential decision-makers: a scheduler and a controller. The scheduler continuously monitors the system's state but transmits it to the controller intermittently to balance the communication cost and control performance. The controller, in turn, determines the control input based on the intermittently received information. Given the partially nested information structure, we show that the optimal control policy follows a certainty-equivalence form. Subsequently, we analyze the qualitative behavior of the scheduling policy. To develop the optimal scheduling policy, we propose InterQ, a deep reinforcement learning algorithm which uses a deep neural network to approximate the Q-function. Through extensive numerical evaluations, we analyze the scheduling landscape and further compare our approach against two baseline strategies: (a) a multi-period periodic scheduling policy, and (b) an event-triggered policy. The results demonstrate that our proposed method outperforms both baselines. The open source implementation can be found at https://github.com/AC-sh/InterQ.
Large Language Models (LLMs) are indispensable in today's applications, but their inference procedure -- generating responses by processing text in segments and using a memory-heavy Key-Value (KV) cache -- demands significant computational resources, particularly under memory constraints. This paper formulates LLM inference optimization as a multi-stage online scheduling problem where sequential prompt arrivals and KV cache growth render conventional scheduling ineffective. We develop a fluid dynamics approximation to provide a tractable benchmark that guides algorithm design. Building on this, we propose the Waiting for Accumulated Inference Threshold (WAIT) algorithm, which uses multiple thresholds to schedule incoming prompts optimally when output lengths are known, and extend it to Nested WAIT for cases with unknown output lengths. Theoretical analysis shows that both algorithms achieve near-optimal performance against the fluid benchmark in heavy traffic conditions, balancing throughput, latency, and Time to First Token (TTFT). Experiments with the Llama-7B model on an A100 GPU using both synthetic and real-world datasets demonstrate improved throughput and reduced latency relative to established baselines like vLLM and Sarathi. This work bridges operations research and machine learning, offering a rigorous framework for the efficient deployment of LLMs under memory constraints.
Large language model (LLM) inference serving systems are essential to various LLM-based applications. As demand for LLM services continues to grow, scaling these systems to handle high request rates while meeting latency Service-Level Objectives (SLOs), referred to as effective throughput, becomes critical. However, existing systems often struggle to improve effective throughput, primarily due to a significant decline in Time To First Token (TTFT) SLO attainment. We identify two major causes of this bottleneck: (1) memory-intensive KV cache that limits batch size expansion under GPU memory constraints, and (2) rigid batch composition enforced by the default First-Come-First-Serve scheduling policy. In this paper, we introduce Apt-Serve, a scalable framework designed to enhance effective throughput in LLM inference serving. Apt-Serve features a new hybrid cache scheme that combines KV cache with a memory-efficient hidden cache for reusable input hidden state vectors, allowing large batch sizes and improving request concurrency. Based on the hybrid cache, Apt-Serve employs an adaptive runtime scheduling mechanism that dynamically optimizes batch composition. We formally define the adaptive scheduling optimization problem and propose an efficient algorithm with theoretical guarantees. Extensive evaluations on three real-world datasets and LLMs ranging from 13B to 66B parameters demonstrate that Apt-Serve achieves up to 8.8x improvement in effective throughput compared to the state-of-the-art inference serving systems.
Traditional 3D content representations include dense point clouds that consume large amounts of data and hence network bandwidth, while newer representations such as neural radiance fields suffer from poor frame rates due to their non-standard volumetric rendering pipeline. 3D Gaussian splats (3DGS) can be seen as a generalization of point clouds that meet the best of both worlds, with high visual quality and efficient rendering for real-time frame rates. However, delivering 3DGS scenes from a hosting server to client devices is still challenging due to high network data consumption (e.g., 1.5 GB for a single scene). The goal of this work is to create an efficient 3D content delivery framework that allows users to view high quality 3D scenes with 3DGS as the underlying data representation. The main contributions of the paper are: (1) Creating new layered 3DGS scenes for efficient delivery, (2) Scheduling algorithms to choose what splats to download at what time, and (3) Trace-driven experiments from users wearing virtual reality headsets to evaluate the visual quality and latency. Our system for Layered 3D Gaussian Splats delivery L3GS demonstrates high visual quality, achieving 16.9% higher average SSIM compared to baselines, and also works with other compressed 3DGS representations.
The Mixture of Experts (MoE) architecture has demonstrated significant advantages as it enables to increase the model capacity without a proportional increase in computation. However, the large MoE model size still introduces substantial memory demands, which usually requires expert offloading on resource-constrained platforms and incurs significant overhead. Hybrid CPU-GPU inference has been proposed to leverage CPU computation to reduce expert loading overhead but faces major challenges: on one hand, the expert activation patterns of MoE models are highly unstable, rendering the fixed mapping strategies in existing works inefficient; on the other hand, the hybrid CPU-GPU schedule for MoE is inherently complex due to the diverse expert sizes, structures, uneven workload distribution, etc. To address these challenges, in this paper, we propose HybriMoE, a hybrid CPU-GPU inference framework that improves resource utilization through a novel CPU-GPU scheduling and cache management system. HybriMoE introduces (i) a dynamic intra-layer scheduling strategy to balance workloads across CPU and GPU, (ii) an impact-driven inter-layer prefetching algorithm, and (iii) a score-based caching algorithm to mitigate expert activation instability. We implement HybriMoE on top of the kTransformers framework and evaluate it on three widely used MoE-based LLMs. Experimental results demonstrate that HybriMoE achieves an average speedup of 1.33$\times$ in the prefill stage and 1.70$\times$ in the decode stage compared to state-of-the-art hybrid MoE inference framework. Our code is available at: https://github.com/PKU-SEC-Lab/HybriMoE.
To address the challenges of limited Battery Swap Stations datasets, high operational costs, and fluctuating user charging demand, this research proposes a probability estimation model based on charging pile data and constructs nine scenario-specific battery swap demand datasets. In addition, this study combines Least Recently Used strategy with Genetic Algorithm and incorporates a guided search mechanism, which effectively enhances the global optimization capability. Thus, a dual-factor decision-making based charging schedule optimization system is constructed. Experimental results show that the constructed datasets exhibit stable trend characteristics, adhering to 24-hour and 168-hour periodicity patterns, with outlier ratios consistently below 3.26%, confirming data validity. Compared to baseline, the improved algorithm achieves better fitness individuals in 80% of test regions under the same iterations. When benchmarked against immediate swap-and-charge strategy, our algorithm achieves a peak cost reduction of 13.96%. Moreover, peak user satisfaction reaches 98.57%, while the average iteration time remains below 0.6 seconds, demonstrating good computational efficiency. The complete datasets and optimization algorithm are open-sourced at https://github.com/qingshufan/GA-EVLRU.
Diffusion probabilistic models (DPMs), while effective in generating high-quality samples, often suffer from high computational costs due to their iterative sampling process. To address this, we propose an enhanced ODE-based sampling method for DPMs inspired by Richardson extrapolation, which reduces numerical error and improves convergence rates. Our method, RX-DPM, leverages multiple ODE solutions at intermediate time steps to extrapolate the denoised prediction in DPMs. This significantly enhances the accuracy of estimations for the final sample while maintaining the number of function evaluations (NFEs). Unlike standard Richardson extrapolation, which assumes uniform discretization of the time grid, we develop a more general formulation tailored to arbitrary time step scheduling, guided by local truncation error derived from a baseline sampling method. The simplicity of our approach facilitates accurate estimation of numerical solutions without significant computational overhead, and allows for seamless and convenient integration into various DPMs and solvers. Additionally, RX-DPM provides explicit error estimates, effectively demonstrating the faster convergence as the leading error term's order increases. Through a series of experiments, we show that the proposed method improves the quality of generated samples without requiring additional sampling iterations.
Video Large Language Models (Video LLMs) have achieved significant success by leveraging a two-stage paradigm: pretraining on large-scale video-text data for vision-language alignment, followed by supervised fine-tuning (SFT) for task-specific capabilities. However, existing approaches struggle with temporal reasoning due to weak temporal correspondence in the data and reliance on the next-token prediction paradigm during training. To address these limitations, we propose TEMPLE (TEMporal Preference Learning), a systematic framework that enhances Video LLMs' temporal reasoning capabilities through Direct Preference Optimization (DPO). To facilitate this, we introduce an automated preference data generation pipeline that systematically constructs preference pairs by selecting videos that are rich in temporal information, designing video-specific perturbation strategies, and finally evaluating model responses on clean and perturbed video inputs. Our temporal alignment features two key innovations: curriculum learning which that progressively increases perturbation difficulty to improve model robustness and adaptability; and "Pre-SFT Alignment'', applying preference optimization before instruction tuning to prioritize fine-grained temporal comprehension. Extensive experiments demonstrate that our approach consistently improves Video LLM performance across multiple benchmarks with a relatively small set of self-generated DPO data. We further analyze the transferability of DPO data across architectures and the role of difficulty scheduling in optimization. Our findings highlight our TEMPLE as a scalable and efficient complement to SFT-based methods, paving the way for developing reliable Video LLMs. Code is available at https://github.com/lscpku/TEMPLE.
Scalable and reproducible policy evaluation has been a long-standing challenge in robot learning. Evaluations are critical to assess progress and build better policies, but evaluation in the real world, especially at a scale that would provide statistically reliable results, is costly in terms of human time and hard to obtain. Evaluation of increasingly generalist robot policies requires an increasingly diverse repertoire of evaluation environments, making the evaluation bottleneck even more pronounced. To make real-world evaluation of robotic policies more practical, we propose AutoEval, a system to autonomously evaluate generalist robot policies around the clock with minimal human intervention. Users interact with AutoEval by submitting evaluation jobs to the AutoEval queue, much like how software jobs are submitted with a cluster scheduling system, and AutoEval will schedule the policies for evaluation within a framework supplying automatic success detection and automatic scene resets. We show that AutoEval can nearly fully eliminate human involvement in the evaluation process, permitting around the clock evaluations, and the evaluation results correspond closely to ground truth evaluations conducted by hand. To facilitate the evaluation of generalist policies in the robotics community, we provide public access to multiple AutoEval scenes in the popular BridgeData robot setup with WidowX robot arms. In the future, we hope that AutoEval scenes can be set up across institutions to form a diverse and distributed evaluation network.
Graph Neural Networks (GNNs) are playing an increasingly important role in the efficient operation and security of computing systems, with applications in workload scheduling, anomaly detection, and resource management. However, their vulnerability to network perturbations poses a significant challenge. We propose $\beta$-GNN, a model enhancing GNN robustness without sacrificing clean data performance. $\beta$-GNN uses a weighted ensemble, combining any GNN with a multi-layer perceptron. A learned dynamic weight, $\beta$, modulates the GNN's contribution. This $\beta$ not only weights GNN influence but also indicates data perturbation levels, enabling proactive mitigation. Experimental results on diverse datasets show $\beta$-GNN's superior adversarial accuracy and attack severity quantification. Crucially, $\beta$-GNN avoids perturbation assumptions, preserving clean data structure and performance.
Mining process optimization particularly truck dispatch scheduling is a critical factor in enhancing the efficiency of open pit mining operations However the dynamic and stochastic nature of mining environments characterized by uncertainties such as equipment failures truck maintenance and variable haul cycle times poses significant challenges for traditional optimization methods While Reinforcement Learning RL has shown promise in adaptive decision making for mining logistics its practical deployment requires rigorous evaluation in realistic and customizable simulation environments The lack of standardized benchmarking environments limits fair algorithm comparisons reproducibility and the real world applicability of RL based approaches in open pit mining settings To address this challenge we introduce Mining Gym a configurable open source benchmarking environment designed for training testing and comparing RL algorithms in mining process optimization Built on Discrete Event Simulation DES and seamlessly integrated with the OpenAI Gym interface Mining Gym provides a structured testbed that enables the direct application of advanced RL algorithms from Stable Baselines The framework models key mining specific uncertainties such as equipment failures queue congestion and the stochasticity of mining processes ensuring a realistic and adaptive learning environment Additionally Mining Gym features a graphical user interface GUI for intuitive mine site configuration a comprehensive data logging system a built in KPI dashboard and real time visual representation of the mine site These capabilities facilitate standardized reproducible evaluations across multiple RL strategies and baseline heuristics
Various parallelism, such as data, tensor, and pipeline parallelism, along with memory optimizations like activation checkpointing, redundancy elimination, and offloading, have been proposed to accelerate distributed training for Large Language Models. To find the best combination of these techniques, automatic distributed training systems are proposed. However, existing systems only tune a subset of optimizations, due to the lack of overlap awareness, inability to navigate the vast search space, and ignoring the inter-microbatch imbalance, leading to sub-optimal performance. To address these shortcomings, we propose Mist, a memory, overlap, and imbalance-aware automatic distributed training system that comprehensively co-optimizes all memory footprint reduction techniques alongside parallelism. Mist is based on three key ideas: (1) fine-grained overlap-centric scheduling, orchestrating optimizations in an overlapped manner, (2) symbolic-based performance analysis that predicts runtime and memory usage using symbolic expressions for fast tuning, and (3) imbalance-aware hierarchical tuning, decoupling the process into an inter-stage imbalance and overlap aware Mixed Integer Linear Programming problem and an intra-stage Dual-Objective Constrained Optimization problem, and connecting them through Pareto frontier sampling. Our evaluation results show that Mist achieves an average of 1.28$\times$ (up to 1.73$\times$) and 1.27$\times$ (up to 2.04$\times$) speedup compared to state-of-the-art manual system Megatron-LM and state-of-the-art automatic system Aceso, respectively.
In large language model (LLM) serving systems, executing each request consists of two phases: the compute-intensive prefill phase and the memory-intensive decoding phase. To prevent performance interference between the two phases, current LLM serving systems typically adopt prefill-decoding disaggregation, where the two phases are split across separate machines. However, we observe this approach leads to significant resource underutilization. Specifically, prefill instances that are compute-intensive suffer from low memory utilization, while decoding instances that are memory-intensive experience low compute utilization. To address this problem, this paper proposes Adrenaline, an attention disaggregation and offloading mechanism designed to enhance resource utilization and performance in LLM serving systems. Adrenaline's key innovation lies in disaggregating part of the attention computation in the decoding phase and offloading them to prefill instances. The memory-bound nature of decoding-phase attention computation inherently enables an effective offloading strategy, yielding two complementary advantages: 1) improved memory capacity and bandwidth utilization in prefill instances, and 2) increased decoding batch sizes that enhance compute utilization in decoding instances, collectively boosting overall system performance. Adrenaline achieves these gains through three key techniques: low-latency decoding synchronization, resource-efficient prefill colocation, and load-aware offloading scheduling. Experimental results show that Adrenaline achieves 2.28x higher memory capacity and 2.07x better memory bandwidth utilization in prefill instances, up to 1.67x improvements in compute utilization for decoding instances, and 1.68x higher overall inference throughput compared to state-of-the-art systems.
Word games hold significant research value for natural language processing (NLP), game theory, and related fields due to their rule-based and situational nature. This study explores how large language models (LLMs) can be effectively involved in word games and proposes a training-free framework. "Shei Shi Wo Di" or "Who is the Spy" in English, is a classic word game. Using this game as an example, we introduce a Chain-of-Thought (CoT)-based scheduling framework to enable LLMs to achieve excellent performance in tasks such as inferring role words and disguising their identities. We evaluate the framework's performance based on game success rates and the accuracy of the LLM agents' analytical results. Experimental results affirm the framework's effectiveness, demonstrating notable improvements in LLM performance across multiple datasets. This work highlights the potential of LLMs in mastering situational reasoning and social interactions within structured game environments. Our code is publicly available at https://github.com/ct-wei/Who-is-The-Spy.
Recent advancements in LLM pretraining have featured ever-expanding context windows to process longer sequences. However, our pilot study reveals that models pretrained with shorter context windows consistently outperform their long-context counterparts under a fixed token budget. This finding motivates us to explore an optimal context window scheduling strategy to better balance long-context capability with pretraining efficiency. To this end, we propose SkyLadder, a simple yet effective approach that implements a short-to-long context window transition. SkyLadder preserves strong standard benchmark performance, while matching or exceeding baseline results on long context tasks. Through extensive experiments, we pre-train 1B-parameter models (up to 32K context) and 3B-parameter models (8K context) on 100B tokens, demonstrating that SkyLadder yields consistent gains of up to 3.7% on common benchmarks, while achieving up to 22% faster training speeds compared to baselines. The code is at https://github.com/sail-sg/SkyLadder.
Bridging different modalities lies at the heart of cross-modality generation. While conventional approaches treat the text modality as a conditioning signal that gradually guides the denoising process from Gaussian noise to the target image modality, we explore a much simpler paradigm-directly evolving between text and image modalities through flow matching. This requires projecting both modalities into a shared latent space, which poses a significant challenge due to their inherently different representations: text is highly semantic and encoded as 1D tokens, whereas images are spatially redundant and represented as 2D latent embeddings. To address this, we introduce FlowTok, a minimal framework that seamlessly flows across text and images by encoding images into a compact 1D token representation. Compared to prior methods, this design reduces the latent space size by 3.3x at an image resolution of 256, eliminating the need for complex conditioning mechanisms or noise scheduling. Moreover, FlowTok naturally extends to image-to-text generation under the same formulation. With its streamlined architecture centered around compact 1D tokens, FlowTok is highly memory-efficient, requires significantly fewer training resources, and achieves much faster sampling speeds-all while delivering performance comparable to state-of-the-art models. Code will be available at https://github.com/bytedance/1d-tokenizer.
The growing demand for computational power is driven by advancements in deep learning, the increasing need for big data processing, and the requirements of scientific simulations for academic and research purposes. Developing countries like Nepal often struggle with the resources needed to invest in new and better hardware for these purposes. However, optimizing and building on existing technology can still meet these computing demands effectively. To address these needs, we built a cluster using four NVIDIA GeForce GTX 1650 GPUs. The cluster consists of four nodes: one master node that controls and manages the entire cluster, and three compute nodes dedicated to processing tasks. The master node is equipped with all necessary software for package management, resource scheduling, and deployment, such as Anaconda and Slurm. In addition, a Network File Storage (NFS) system was integrated to provide the additional storage required by the cluster. Given that the cluster is accessible via ssh by a public domain address, which poses significant cybersecurity risks, we implemented fail2ban to mitigate brute force attacks and enhance security. Despite the continuous challenges encountered during the design and implementation process, this project demonstrates how powerful computational clusters can be built to handle resource-intensive tasks in various demanding fields.
Recent advancements in probing Large Language Models (LLMs) have explored their latent potential as personalized travel planning agents, yet existing benchmarks remain limited in real world applicability. Existing datasets, such as TravelPlanner and TravelPlanner+, suffer from semi synthetic data reliance, spatial inconsistencies, and a lack of key travel constraints, making them inadequate for practical itinerary generation. To address these gaps, we introduce TripCraft, a spatiotemporally coherent travel planning dataset that integrates real world constraints, including public transit schedules, event availability, diverse attraction categories, and user personas for enhanced personalization. To evaluate LLM generated plans beyond existing binary validation methods, we propose five continuous evaluation metrics, namely Temporal Meal Score, Temporal Attraction Score, Spatial Score, Ordering Score, and Persona Score which assess itinerary quality across multiple dimensions. Our parameter informed setting significantly enhances meal scheduling, improving the Temporal Meal Score from 61% to 80% in a 7 day scenario. TripCraft establishes a new benchmark for LLM driven personalized travel planning, offering a more realistic, constraint aware framework for itinerary generation. Dataset and Codebase will be made publicly available upon acceptance.
Large Language Models (LLMs) have shown remarkable capabilities across various domains, but their potential for solving combinatorial optimization problems remains largely unexplored. In this paper, we investigate the applicability of LLMs to the Job Shop Scheduling Problem (JSSP), a classic challenge in combinatorial optimization that requires efficient job allocation to machines to minimize makespan. To this end, we introduce Starjob, the first supervised dataset for JSSP, comprising 130k instances specifically designed for training LLMs. Leveraging this dataset, we fine-tune the LLaMA 8B 4-bit quantized model with the LoRA method to develop an end-to-end scheduling approach. Our evaluation on standard benchmarks demonstrates that the proposed LLM-based method not only surpasses traditional Priority Dispatching Rules (PDRs) but also achieves notable improvements over state-of-the-art neural approaches like L2D, with an average improvement of 15.36% on DMU and 7.85% on Taillard benchmarks. These results highlight the untapped potential of LLMs in tackling combinatorial optimization problems, paving the way for future advancements in this area.
Federated Learning (FL) is an approach for privacy-preserving Machine Learning (ML), enabling model training across multiple clients without centralized data collection. With an aggregator server coordinating training, aggregating model updates, and storing metadata across rounds. In addition to training, a substantial part of FL systems are the non-training workloads such as scheduling, personalization, clustering, debugging, and incentivization. Most existing systems rely on the aggregator to handle non-training workloads and use cloud services for data storage. This results in high latency and increased costs as non-training workloads rely on large volumes of metadata, including weight parameters from client updates, hyperparameters, and aggregated updates across rounds, making the situation even worse. We propose FLStore, a serverless framework for efficient FL non-training workloads and storage. FLStore unifies the data and compute planes on a serverless cache, enabling locality-aware execution via tailored caching policies to reduce latency and costs. Per our evaluations, compared to cloud object store based aggregator server FLStore reduces per request average latency by 71% and costs by 92.45%, with peak improvements of 99.7% and 98.8%, respectively. Compared to an in-memory cloud cache based aggregator server, FLStore reduces average latency by 64.6% and costs by 98.83%, with peak improvements of 98.8% and 99.6%, respectively. FLStore integrates seamlessly with existing FL frameworks with minimal modifications, while also being fault-tolerant and highly scalable.