arXiv subjects

Jingyuan Cheng

Publications and source records attributed to Jingyuan Cheng.

5 recordsLinked to original sources

arXiv2026

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models -- DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) -- both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) Manifold-Constrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability. We pre-train both models on more than 32T diverse and high-quality tokens, followed by a comprehensive post-training pipeline that unlocks and further enhances their capabilities. DeepSeek-V4-Pro-Max, the maximum reasoning effort mode of DeepSeek-V4-Pro, redefines the state-of-the-art for open models, outperforming its predecessors in core tasks. Meanwhile, DeepSeek-V4 series are highly efficient in long-context scenarios. In the one-million-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek-V3.2. This enables us to routinely support one-million-token contexts, thereby making long-horizon tasks and further test-time scaling more feasible. The model checkpoints are available at https://huggingface.co/collections/deepseek-ai/deepseek-v4.

cs.CL↗

arXiv2025

DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). (3) Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This methodology facilitates scalable agentic post-training, yielding substantial improvements in generalization and instruction-following robustness within complex, interactive environments.

cs.CL↗

arXiv2025

Optimization of An Induced Magnetic Field-Based Positioning System

Using oscillating magnetic fields for indoor positioning is a robust way to resist dynamic environments. This work presents the hard- and software-related optimizations of an induced magnetic field positioning system. We describe a new coil architecture for both the transmitter and receiver, reducing inter-axes cross-talk. A new analog circuit design on the receiver side attains an acceptable noise level and increases the detection range from 4m to 8m (the covered area is increased from $50m^2$ to $200m^2$). The median positioning error is reduced from 0.56~m to 0.25m in the near field with fingerprinting methods. Experiments in office and factory areas (including robotic and industrial equipment) demonstrate the system's robustness in large areas. This work aims to enlighten researchers working on the same topic with constructive optimization directions on their own induced magnetic field-based systems.

eess.SP↗

arXiv2023

Dual-modality Smart Shoes for Quantitative Assessment of Hemiplegic Patients' Lower Limbs' Muscle Strength

Stroke can lead to the impaired motor ability of the patient's lower limbs and hemiplegia. Accurate assessment of the lower limbs' motor ability is important for diagnosis and rehabilitation. To digitalize such assessment so that each test can be traced back any time and subjectivity can be avoided, we test how dual-modality smart shoes equipped with pressure-sensitive insoles and inertial measurement units can be used for this purpose. A 5m walking test protocol, including the left and right turns, is designed. Data are collected from 23 patients and 17 healthy subjects. For the lower limbs' motor ability, the tests are observed by two physicians and assessed using the five graded Medical Research Council scale for muscle examination. The average of two physicians' scores for the same patient is used as the ground truth. Using the feature set we developed, 100\% accuracy is achieved in classifying the patients and healthy subjects. For patients' muscle strength, a mean absolute error of 0.143 and a maximum error of 0.395 is achieved using our feature set and the regression method, closer to the ground truth than the scores from each physician (mean absolute error: 0.217, maximum error: 0.5). We thus validate the possibility of using such smart shoes to objectively and accurately evaluate the lower limbs' muscle strength of the stroke patients.

cs.HC↗

arXiv2023

MassNet: A Deep Learning Approach for Body Weight Extraction from A Single Pressure Image

Body weight, as an essential physiological trait, is of considerable significance in many applications like body management, rehabilitation, and drug dosing for patient-specific treatments. Previous works on the body weight estimation task are mainly vision-based, using 2D/3D, depth, or infrared images, facing problems in illumination, occlusions, and especially privacy issues. The pressure mapping mattress is a non-invasive and privacy-preserving tool to obtain the pressure distribution image over the bed surface, which strongly correlates with the body weight of the lying person. To extract the body weight from this image, we propose a deep learning-based model, including a dual-branch network to extract the deep features and pose features respectively. A contrastive learning module is also combined with the deep-feature branch to help mine the mutual factors across different postures of every single subject. The two groups of features are then concatenated for the body weight regression task. To test the model's performance over different hardware and posture settings, we create a pressure image dataset of 10 subjects and 23 postures, using a self-made pressure-sensing bedsheet. This dataset, which is made public together with this paper, together with a public dataset, are used for the validation. The results show that our model outperforms the state-of-the-art algorithms over both 2 datasets. Our research constitutes an important step toward fully automatic weight estimation in both clinical and at-home practice. Our dataset is available for research purposes at: https://github.com/USTCWzy/MassEstimation.

cs.HC↗