Pushing the Limits of Low-Bit Optimizers: A Focus on EMA Dynamics

The explosion in model sizes leads to continued growth in prohibitive training/fine-tuning costs, particularly for stateful optimizers which maintain auxiliary information of even 2x the model size to achieve optimal convergence. We therefore present in this work a novel type of optimizer that carries with extremely lightweight state overloads, achieved through ultra-low-precision quantization. While previous efforts have achieved certain success with 8-bit or 4-bit quantization, our approach enables optimizers to operate at precision as low as 3 bits, or even 2 bits per state element. This is accomplished by identifying and addressing two critical challenges: the signal swamping problem in unsigned quantization that results in unchanged state dynamics, and the rapidly increased gradient variance in signed quantization that leads to incorrect descent directions. The theoretical analysis suggests a tailored logarithmic quantization for the former and a precision-specific momentum value for the latter. Consequently, the proposed SOLO achieves substantial memory savings (approximately 45 GB when training a 7B model) with minimal accuracy loss. We hope that SOLO can contribute to overcoming the bottleneck in computational resources, thereby promoting greater accessibility in fundamental research.

Tempo: Application-aware LLM Serving with Mixed SLO Requirements

The integration of Large Language Models (LLMs) into diverse applications, ranging from interactive chatbots and cloud AIOps to intelligent agents, has introduced a wide spectrum of Service Level Objectives (SLOs) for responsiveness. These workloads include latency-sensitive requests focused on per-token latency in streaming chat, throughput-intensive requests that require rapid full responses to invoke tools, and collective requests with dynamic dependencies arising from self-reflection or agent-based reasoning. This workload diversity, amplified by unpredictable request information such as response lengths and runtime dependencies, makes existing schedulers inadequate even within their design envelopes. In this paper, we define service gain as the useful service delivered by completing requests. We observe that as SLO directly reflects the actual performance needs of requests, completing a request much faster than its SLO (e.g., deadline) yields limited additional service gain. Based on this insight, we introduce Tempo, the first systematic SLO-aware scheduler designed to maximize service gain across diverse LLM workloads. Tempo allocates just enough serving bandwidth to meet each SLO, maximizing residual capacity for others best-effort workloads. Instead of assuming request information or none at all, it adopts a hybrid scheduling strategy: using quantile-based response upper bounds and dependency-graph matching for conservative initial estimates, prioritizing requests by service gain density, and refining decisions online as generation progresses. Our evaluation across diverse workloads, including chat, reasoning, and agentic pipelines, shows that Tempo improves end-to-end service gain by up to 8.3$\times$ and achieves up to 10.3$\times$ SLO goodput compared to state-of-the-art designs

Joint Topology and Power Optimization for Multi-UAV Collaborative Secure Communication

In this paper, we investigate an unmanned aerial vehicle (UAV)-enabled secure communication scenario that a cluster of UAVs performs a virtual non-uniform linear array (NULA) to communicate with a base station (BS) in the presence of eavesdroppers (Eves). Our goal is to design the UAV topology, trajectory, and precoding to maximize the system channel capacity. To this end, we convert the original problem into equivalent two-stage problems. Specifically, we first try to maximize the channel gain by meticulously designing the UAV topology. We then study the joint optimization of the trajectory and precoding for total transmit power minimization while satisfying the constraints on providing quality of service (QoS) assurance to the BS, the leakage tolerance to Eves, the per-UAV transmit power, the initial/final locations, and the cylindrical no-fly zones. For the UAV topology design, we prove that the topology follows the Fekete-point distribution. The design of trajectory and precoding is formulated as a non-convex optimization problem which is generally intractable. Subsequently, the non-convex constraints are converted into convex terms, and a double-loop search algorithm is proposed to solve the transmit power minimization problem. Introduce random rotation offsets so as to perform a dynamic stochastic channel to enhance the security. Numerical results demonstrate the superiority of the proposed method in promoting capacity.

The PHD/CPHD filter for Multiple Extended Target Tracking with Trajectory Set Theory and Explicit Shape Estimation

In this paper, we propose two methods for tracking multiple extended targets or unresolved group targets with elliptical extent shape. These two methods are deduced from the famous Probability Hypothesis Density (PHD) filter and the Cardinality-PHD (CPHD) filter, respectively. In these two methods, Trajectory Set Theory (TST) is combined to establish the target trajectory estimates. Moreover, by employing a decoupled shape estimation model, the proposed methods can explicitly provide the shape estimation of the target, such as the orientation of the ellipse extension and the length of its two axes. We derived the closed Bayesian recursive of these two methods with stable trajectory generation and accurate extent estimation, resulting in the TPHD-E filter and the TCPHD-E filter. In addition, Gaussian mixture implementations of our methods are provided, which are further referred to as the GM-TPHD-E filter and the GM-TCPHD-E filters. We illustrate the ability of these methods through simulations and experiments with real data. These experiments demonstrate that the two proposed algorithms have advantages over existing algorithms in target shape estimation, as well as in the completeness and accuracy of target trajectory generation.

Knowledge Distillation and Dataset Distillation of Large Language Models: Emerging Trends, Challenges, and Future Directions

The exponential growth of Large Language Models (LLMs) continues to highlight the need for efficient strategies to meet ever-expanding computational and data demands. This survey provides a comprehensive analysis of two complementary paradigms: Knowledge Distillation (KD) and Dataset Distillation (DD), both aimed at compressing LLMs while preserving their advanced reasoning capabilities and linguistic diversity. We first examine key methodologies in KD, such as task-specific alignment, rationale-based training, and multi-teacher frameworks, alongside DD techniques that synthesize compact, high-impact datasets through optimization-based gradient matching, latent space regularization, and generative synthesis. Building on these foundations, we explore how integrating KD and DD can produce more effective and scalable compression strategies. Together, these approaches address persistent challenges in model scalability, architectural heterogeneity, and the preservation of emergent LLM abilities. We further highlight applications across domains such as healthcare and education, where distillation enables efficient deployment without sacrificing performance. Despite substantial progress, open challenges remain in preserving emergent reasoning and linguistic diversity, enabling efficient adaptation to continually evolving teacher models and datasets, and establishing comprehensive evaluation protocols. By synthesizing methodological innovations, theoretical foundations, and practical insights, our survey charts a path toward sustainable, resource-efficient LLMs through the tighter integration of KD and DD principles.

SlimPipe: Memory-Thrifty and Efficient Pipeline Parallelism for Long-Context LLM Training

Pipeline Parallelism (PP) serves as a crucial technique for training Large Language Models (LLMs), owing to its capability to alleviate memory pressure from model states with relatively low communication overhead. However, in long-context scenarios, existing pipeline parallelism methods fail to address the substantial activation memory pressure, primarily due to the peak memory consumption resulting from the accumulation of activations across multiple microbatches. Moreover, these approaches inevitably introduce considerable pipeline bubbles, further hindering efficiency. To tackle these challenges, we propose SlimPipe, a novel approach to fine-grained pipeline parallelism that employs uniform sequence slicing coupled with one-forward-one-backward (1F1B) schedule. It reduces the accumulated activations from several microbatches to just one, which is split into several slices. Although the slices are evenly partitioned, the computation cost is not equal across slices due to causal attention. We develop a sophisticated workload redistribution technique to address this load imbalance. SlimPipe achieves (1) near-zero memory overhead and (2) minimal pipeline bubbles simultaneously. The effectiveness of SlimPipe has been proven by thorough testing with diverse model architectures, context window sizes, and SlimPipe-specific configurations. For example, on the Llama 70B model, compared to state-of-the-art methods, SlimPipe significantly boosts the Model FLOPs Utilization (MFU) to up to $1.57\times$ for a context length of 512K. More notably, for a context length of 2048K, it maintains over 45% utilization on 256 NVIDIA Hopper 80GB GPUs, while other approaches either suffer significant performance drops or fail entirely due to memory constraints.

EarthGPT-X: Enabling MLLMs to Flexibly and Comprehensively Understand Multi-Source Remote Sensing Imagery

Recent advances in the visual-language area have developed natural multi-modal large language models (MLLMs) for spatial reasoning through visual prompting. However, due to remote sensing (RS) imagery containing abundant geospatial information that differs from natural images, it is challenging to effectively adapt natural spatial models to the RS domain. Moreover, current RS MLLMs are limited in overly narrow interpretation levels and interaction manner, hindering their applicability in real-world scenarios. To address those challenges, a spatial MLLM named EarthGPT-X is proposed, enabling a comprehensive understanding of multi-source RS imagery, such as optical, synthetic aperture radar (SAR), and infrared. EarthGPT-X offers zoom-in and zoom-out insight, and possesses flexible multi-grained interactive abilities. Moreover, EarthGPT-X unifies two types of critical spatial tasks (i.e., referring and grounding) into a visual prompting framework. To achieve these versatile capabilities, several key strategies are developed. The first is the multi-modal content integration method, which enhances the interplay between images, visual prompts, and text instructions. Subsequently, a cross-domain one-stage fusion training strategy is proposed, utilizing the large language model (LLM) as a unified interface for multi-source multi-task learning. Furthermore, by incorporating a pixel perception module, the referring and grounding tasks are seamlessly unified within a single framework. In addition, the experiments conducted demonstrate the superiority of the proposed EarthGPT-X in multi-grained tasks and its impressive flexibility in multi-modal interaction, revealing significant advancements of MLLM in the RS field.

Dynamic Compressing Prompts for Efficient Inference of Large Language Models

Large Language Models (LLMs) have shown outstanding performance across a variety of tasks, partly due to advanced prompting techniques. However, these techniques often require lengthy prompts, which increase computational costs and can hinder performance because of the limited context windows of LLMs. While prompt compression is a straightforward solution, existing methods confront the challenges of retaining essential information, adapting to context changes, and remaining effective across different tasks. To tackle these issues, we propose a task-agnostic method called Dynamic Compressing Prompts (LLM-DCP). Our method reduces the number of prompt tokens while aiming to preserve the performance as much as possible. We model prompt compression as a Markov Decision Process (MDP), enabling the DCP-Agent to sequentially remove redundant tokens by adapting to dynamic contexts and retaining crucial content. We develop a reward function for training the DCP-Agent that balances the compression rate, the quality of the LLM output, and the retention of key information. This allows for prompt token reduction without needing an external black-box LLM. Inspired by the progressive difficulty adjustment in curriculum learning, we introduce a Hierarchical Prompt Compression (HPC) training strategy that gradually increases the compression difficulty, enabling the DCP-Agent to learn an effective compression method that maintains information integrity. Experiments demonstrate that our method outperforms state-of-the-art techniques, especially at higher compression rates. The code for our approach will be available at https://github.com/Fhujinwu/DCP.

CliniChat: A Multi-Source Knowledge-Driven Framework for Clinical Interview Dialogue Reconstruction and Evaluation

Large language models (LLMs) hold great promise for assisting clinical interviews due to their fluent interactive capabilities and extensive medical knowledge. However, the lack of high-quality interview dialogue data and widely accepted evaluation methods has significantly impeded this process. So we propose CliniChat, a framework that integrates multi-source knowledge to enable LLMs to simulate real-world clinical interviews. It consists of two modules: Clini-Recon and Clini-Eval, each responsible for reconstructing and evaluating interview dialogues, respectively. By incorporating three sources of knowledge, Clini-Recon transforms clinical notes into systematic, professional, and empathetic interview dialogues. Clini-Eval combines a comprehensive evaluation metric system with a two-phase automatic evaluation approach, enabling LLMs to assess interview performance like experts. We contribute MedQA-Dialog, a high-quality synthetic interview dialogue dataset, and CliniChatGLM, a model specialized for clinical interviews. Experimental results demonstrate that CliniChatGLM's interview capabilities undergo a comprehensive upgrade, particularly in history-taking, achieving state-of-the-art performance.

Debiasing 6-DOF IMU via Hierarchical Learning of Continuous Bias Dynamics

This paper develops a deep learning approach to the online debiasing of IMU gyroscopes and accelerometers. Most existing methods rely on implicitly learning a bias term to compensate for raw IMU data. Explicit bias learning has recently shown its potential as a more interpretable and motion-independent alternative. However, it remains underexplored and faces challenges, particularly the need for ground truth bias data, which is rarely available. To address this, we propose a neural ordinary differential equation (NODE) framework that explicitly models continuous bias dynamics, requiring only pose ground truth, often available in datasets. This is achieved by extending the canonical NODE framework to the matrix Lie group for IMU kinematics with a hierarchical training strategy. The validation on two public datasets and one real-world experiment demonstrates significant accuracy improvements in IMU measurements, reducing errors in both pure IMU integration and visual-inertial odometry.