Holistic Capability Preservation: Towards Compact Yet Comprehensive Reasoning Models

This technical report presents Ring-Lite-Distill, a lightweight reasoning model derived from our open-source Mixture-of-Experts (MoE) Large Language Models (LLMs) Ling-Lite. This study demonstrates that through meticulous high-quality data curation and ingenious training paradigms, the compact MoE model Ling-Lite can be further trained to achieve exceptional reasoning capabilities, while maintaining its parameter-efficient architecture with only 2.75 billion activated parameters, establishing an efficient lightweight reasoning architecture. In particular, in constructing this model, we have not merely focused on enhancing advanced reasoning capabilities, exemplified by high-difficulty mathematical problem solving, but rather aimed to develop a reasoning model with more comprehensive competency coverage. Our approach ensures coverage across reasoning tasks of varying difficulty levels while preserving generic capabilities, such as instruction following, tool use, and knowledge retention. We show that, Ring-Lite-Distill's reasoning ability reaches a level comparable to DeepSeek-R1-Distill-Qwen-7B, while its general capabilities significantly surpass those of DeepSeek-R1-Distill-Qwen-7B. The models are accessible at https://huggingface.co/inclusionAI

HybriMoE: Hybrid CPU-GPU Scheduling and Cache Management for Efficient MoE Inference

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.

CDKFormer: Contextual Deviation Knowledge-Based Transformer for Long-Tail Trajectory Prediction

Predicting the future movements of surrounding vehicles is essential for ensuring the safe operation and efficient navigation of autonomous vehicles (AVs) in urban traffic environments. Existing vehicle trajectory prediction methods primarily focus on improving overall performance, yet they struggle to address long-tail scenarios effectively. This limitation often leads to poor predictions in rare cases, significantly increasing the risk of safety incidents. Taking Argoverse 2 motion forecasting dataset as an example, we first investigate the long-tail characteristics in trajectory samples from two perspectives, individual motion and group interaction, and deriving deviation features to distinguish abnormal from regular scenarios. On this basis, we propose CDKFormer, a Contextual Deviation Knowledge-based Transformer model for long-tail trajectory prediction. CDKFormer integrates an attention-based scene context fusion module to encode spatiotemporal interaction and road topology. An additional deviation feature fusion module is proposed to capture the dynamic deviations in the target vehicle status. We further introduce a dual query-based decoder, supported by a multi-stream decoder block, to sequentially decode heterogeneous scene deviation features and generate multimodal trajectory predictions. Extensive experiments demonstrate that CDKFormer achieves state-of-the-art performance, significantly enhancing prediction accuracy and robustness for long-tailed trajectories compared to existing methods, thus advancing the reliability of AVs in complex real-world environments.

Deep Learning-Powered Electrical Brain Signals Analysis: Advancing Neurological Diagnostics

Neurological disorders represent significant global health challenges, driving the advancement of brain signal analysis methods. Scalp electroencephalography (EEG) and intracranial electroencephalography (iEEG) are widely used to diagnose and monitor neurological conditions. However, dataset heterogeneity and task variations pose challenges in developing robust deep learning solutions. This review systematically examines recent advances in deep learning approaches for EEG/iEEG-based neurological diagnostics, focusing on applications across 7 neurological conditions using 46 datasets. We explore trends in data utilization, model design, and task-specific adaptations, highlighting the importance of pre-trained multi-task models for scalable, generalizable solutions. To advance research, we propose a standardized benchmark for evaluating models across diverse datasets to enhance reproducibility. This survey emphasizes how recent innovations can transform neurological diagnostics and enable the development of intelligent, adaptable healthcare solutions.

LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-design

State space models (SSMs) like Mamba have recently attracted much attention. Compared to Transformer-based large language models (LLMs), Mamba achieves linear computation complexity with the sequence length and demonstrates superior performance. However, Mamba is hard to accelerate due to the scattered activation outliers and the complex computation dependency, rendering existing LLM accelerators inefficient. In this paper, we propose LightMamba that co-designs the quantization algorithm and FPGA accelerator architecture for efficient Mamba inference. We first propose an FPGA-friendly post-training quantization algorithm that features rotation-assisted quantization and power-of-two SSM quantization to reduce the majority of computation to 4-bit. We further design an FPGA accelerator that partially unrolls the Mamba computation to balance the efficiency and hardware costs. Through computation reordering as well as fine-grained tiling and fusion, the hardware utilization and memory efficiency of the accelerator get drastically improved. We implement LightMamba on Xilinx Versal VCK190 FPGA and achieve 4.65x to 6.06x higher energy efficiency over the GPU baseline. When evaluated on Alveo U280 FPGA, LightMamba reaches 93 tokens/s, which is 1.43x that of the GPU baseline.

SMamba: Sparse Mamba for Event-based Object Detection

Transformer-based methods have achieved remarkable performance in event-based object detection, owing to the global modeling ability. However, they neglect the influence of non-event and noisy regions and process them uniformly, leading to high computational overhead. To mitigate computation cost, some researchers propose window attention based sparsification strategies to discard unimportant regions, which sacrifices the global modeling ability and results in suboptimal performance. To achieve better trade-off between accuracy and efficiency, we propose Sparse Mamba (SMamba), which performs adaptive sparsification to reduce computational effort while maintaining global modeling capability. Specifically, a Spatio-Temporal Continuity Assessment module is proposed to measure the information content of tokens and discard uninformative ones by leveraging the spatiotemporal distribution differences between activity and noise events. Based on the assessment results, an Information-Prioritized Local Scan strategy is designed to shorten the scan distance between high-information tokens, facilitating interactions among them in the spatial dimension. Furthermore, to extend the global interaction from 2D space to 3D representations, a Global Channel Interaction module is proposed to aggregate channel information from a global spatial perspective. Results on three datasets (Gen1, 1Mpx, and eTram) demonstrate that our model outperforms other methods in both performance and efficiency.

Data and System Perspectives of Sustainable Artificial Intelligence

Sustainable AI is a subfield of AI for concerning developing and using AI systems in ways of aiming to reduce environmental impact and achieve sustainability. Sustainable AI is increasingly important given that training of and inference with AI models such as large langrage models are consuming a large amount of computing power. In this article, we discuss current issues, opportunities and example solutions for addressing these issues, and future challenges to tackle, from the data and system perspectives, related to data acquisition, data processing, and AI model training and inference.

DeepSeek-V3 Technical Report

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks. The model checkpoints are available at https://github.com/deepseek-ai/DeepSeek-V3.

RoboMIND: Benchmark on Multi-embodiment Intelligence Normative Data for Robot Manipulation

Developing robust and general-purpose robotic manipulation policies is a key goal in the field of robotics. To achieve effective generalization, it is essential to construct comprehensive datasets that encompass a large number of demonstration trajectories and diverse tasks. Unlike vision or language data that can be collected from the Internet, robotic datasets require detailed observations and manipulation actions, necessitating significant investment in hardware-software infrastructure and human labor. While existing works have focused on assembling various individual robot datasets, there remains a lack of a unified data collection standard and insufficient diversity in tasks, scenarios, and robot types. In this paper, we introduce RoboMIND (Multi-embodiment Intelligence Normative Data for Robot manipulation), featuring 55k real-world demonstration trajectories across 279 diverse tasks involving 61 different object classes. RoboMIND is collected through human teleoperation and encompasses comprehensive robotic-related information, including multi-view RGB-D images, proprioceptive robot state information, end effector details, and linguistic task descriptions. To ensure dataset consistency and reliability during policy learning, RoboMIND is built on a unified data collection platform and standardized protocol, covering four distinct robotic embodiments. We provide a thorough quantitative and qualitative analysis of RoboMIND across multiple dimensions, offering detailed insights into the diversity of our datasets. In our experiments, we conduct extensive real-world testing with four state-of-the-art imitation learning methods, demonstrating that training with RoboMIND data results in a high manipulation success rate and strong generalization. Our project is at https://x-humanoid-robomind.github.io/.

Learning-based Sketches for Frequency Estimation in Data Streams without Ground Truth

Estimating the frequency of items on the high-volume, fast data stream has been extensively studied in many areas, such as database and network measurement. Traditional sketch algorithms only allow to give very rough estimates with limited memory cost, whereas some learning-augmented algorithms have been proposed recently, their offline framework requires actual frequencies that are challenging to access in general for training, and speed is too slow for real-time processing, despite the still coarse-grained accuracy. To this end, we propose a more practical learning-based estimation framework namely UCL-sketch, by following the line of equation-based sketch to estimate per-key frequencies. In a nutshell, there are two key techniques: online training via equivalent learning without ground truth, and highly scalable architecture with logical estimation buckets. We implemented experiments on both real-world and synthetic datasets. The results demonstrate that our method greatly outperforms existing state-of-the-art sketches regarding per-key accuracy and distribution, while preserving resource efficiency. Our code is attached in the supplementary material, and will be made publicly available at https://github.com/Y-debug-sys/UCL-sketch.