Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks99 from public data
- Seedream 4.0: Toward Next-generation Multimodal Image Generation253
This work introduces Seedream 4.0, an efficient and high-performance multimodal image generation system that unifies text-to-image (T2I) synthesis, image editing, and multi-image composition within a single framework, and develops a highly efficient diffusion transformer with a powerful VAE which allows for efficient training of the model.
- VmambaIR: Visual State Space Model for Image Restoration242
The proposed VmambaIR is one of the first works to introduce State Space Models (SSMs) with linear complexity into comprehensive image restoration tasks, and highlights the potential of state space models as promising alternatives to the transformer and CNN architectures in serving as foundational frameworks for next-generation low-level visual tasks.
- Control-A-Video: Controllable Text-to-Video Diffusion Models with Motion Prior and Reward Feedback Learning240
A controllable text-to-video (T2V) diffusion model, called Control-A-Video, capable of maintaining consistency while customizable video synthesis, based on a pre-trained conditional text-to-image (T2I) diffusion model.
- Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarios235
Extensive experiments show that Next-ViT significantly outperforms existing CNNs, ViTs and CNN-Transformer hybrid architectures with respect to the latency/accuracy trade-off across various vision tasks.
- Building fast and compact convolutional neural networks for offline handwritten Chinese character recognition165
A nine-layer CNN for HCCR consisting of 3755 classes is designed, and an algorithm is devised that can reduce the networks computational cost by nine times and compress the network to 1/18 of the original size of the baseline model, with only a 0.21% drop in accuracy.
- FreeSeg: Unified, Universal and Open-Vocabulary Image Segmentation156
FreeSeg is proposed, a generic framework to accomplish Unified, Universal and Open-Vocabulary Image Segmentation, which optimizes an all-in-one network via one-shot training and employs the same architecture and parameters to handle diverse segmentation tasks seamlessly in the inference procedure.
- Diffusion Adversarial Post-Training for One-Step Video Generation153
This work proposes Adversarial Post-Training (APT) against real data following diffusion pre-training for one-step video generation, and introduces several improvements to the model architecture and training procedures, along with an approximated R1 regularization objective.
- Seedream 3.0 Technical Report150
Overall capabilities of Seedream 3.0 are enhanced, in particular for text-rendering in complicated Chinese characters which is important to professional typography generation, and it provides native high-resolution output, allowing it to generate images with high visual quality.
- 149
- Activation Modulation and Recalibration Scheme for Weakly Supervised Semantic Segmentation146
A novel activation modulation and recalibration scheme that establishes a new state-of-the-art performance on the PASCAL VOC 2012 dataset, surpassing not only current methods trained with the image-level of supervision but also some methods relying on stronger supervision, such as saliency label.
- AutoDiffusion: Training-Free Optimization of Time Steps and Architectures for Automated Diffusion Model Acceleration94
This work designs a unified search space that consists of all possible time steps and various architectures and introduces a two stage evolutionary algorithm to find the optimal solution in the designed search space.
- Seaweed-7B: Cost-Effective Training of Video Generation Foundation Model90
This technical report presents a cost-efficient strategy for training a video generation foundation model called Seaweed-7B trained from scratch using 665,000 H100 GPU hours and highlights the key design decisions that enhance the performance of the medium-sized diffusion model.
- 81
- Seedance 1.5 pro: A Native Audio-Visual Joint Generation Foundation Model73
This work introduces an acceleration framework that boosts inference speed by over 10X and distinguishes itself through precise multilingual and dialect lip-syncing, dynamic cinematic camera control, and enhanced narrative coherence, positioning it as a robust engine for professional-grade content creation.
- Heavy metal removal from aqueous solutions by chitosan-based magnetic composite flocculants73
Characterization results show that the three magnetic carboxymethyl chitosan-based flocculant have been successfully prepared with good magnetic induction properties and shows excellent flocculation performance in removing soluble heavy metals.
- AffineQuant: Affine Transformation Quantization for Large Language Models72
This paper advocates for the direct optimization using equivalent Affine transformations in PTQ (AffineQuant) using the corresponding inverse matrix to ensure equivalence between the pre- and post-quantization outputs of PTQ, thereby maintaining its efficiency and generalization capabilities.
- Thermosensitive in situ nanogel as ophthalmic delivery system of curcumin: development, characterization, in vitro permeation and in vivo pharmacokinetic studies71
Results demonstrated that CUR-CNLC-GEL could become a potential formulation for increasing the bioavailability of CUR in the aqueous humor by enhancing corneal permeation and retention capacity.
- Seedream 2.0: A Native Chinese-English Bilingual Image Generation Foundation Model67
This work presents Seedream 2.0, a native Chinese-English bilingual image generation foundation model that excels across diverse dimensions, which adeptly manages text prompt in both Chinese and English, supporting bilingual image generation and text rendering.
- 67
- 61
- Flocculation of combined contaminants of dye and heavy metal by nano-chitosan flocculants60
It is proved that nanoflocculants made by modifying CMCTS with different structures can demonstrate ideal flocculation removal performance for dye and heavy metal wastewaters.
- 60
- SepViT: Separable Vision Transformer55
The novel window token embedding and grouped self-attention are employed to compute the attention relationship among windows with negligible cost and establish long-range visual interactions across multiple windows, respectively.
- 54
- Online Multi-Granularity Distillation for GAN Compression54
This work proposes a novel online multi-granularity distillation (OMGD) scheme to obtain lightweight GANs, which contributes to generating highfidelity images with low computational demands and reveals that OMGD provides a feasible solution for the deployment of real-time image translation on resource-constrained devices.
- SeedEdit 3.0: Fast and High-Quality Generative Image Editing53
An enhanced data curation pipeline with a meta-info paradigm and meta-info embedding strategy that help mix images from multiple data sources is developed, which allows to scale editing data effectively and connect VLM with diffusion model more closely.
- SeedVR2: One-Step Video Restoration via Diffusion Adversarial Post-Training51
This work proposes a one-step diffusion-based VR model, termed as SeedVR2, which performs adversarial VR training against real data, and introduces several enhancements to both model architecture and training procedures to handle the challenging high-resolution VR within a single step.
- 50
- 47
- Adversarial Distribution Matching for Diffusion Distillation Towards Efficient Image and Video Synthesis43
This work proposes Adversarial Distribution Matching (ADM), a novel framework that leverages diffusion-based discriminators to align the latent predictions between real and fake score estimators for score distillation in an adversarial manner and achieves superior one-step performance on SDXL compared to DMD2 while consuming less GPU time.
- DiffusionEngine: Diffusion Model is Scalable Data Engine for Object Detection43
This paper reveals that the recently developed Diffusion Model is a scalable data engine for object detection, consisting of a pre-trained diffusion model and an effective Detection-Adapter, contributing to generating scalable, diverse and generalizable detection data in a plug-and-play manner.
- FlexSP: Accelerating Large Language Model Training via Flexible Sequence Parallelism41
This paper proposes a heterogeneity-adaptive sequence parallelism method that captures the variability in sequence lengths and assigns the optimal combination of scattering strategies based on workload characteristics and implements this method in a high-performance system that supports adaptive parallelization in distributed LLM training.
- AlignDet: Aligning Pre-training and Fine-tuning in Object Detection40
AlignDet is proposed, a unified pre-training framework that can be adapted to various existing detectors to alleviate discrepancies in data, model, and task between the pre-training and fine-tuning procedure in existing practices, which implicitly limit the detector’s performance, generalization ability, and convergence speed.
- TRT-ViT: TensorRT-oriented Vision Transformer36
This paper directly treats the TensorRT latency on the specific hardware as an efficiency metric, which provides more comprehensive feedback involving computational capacity, memory cost, and bandwidth, and a family of TensortRT-oriented Transformers, abbreviated as TRT-ViT.
- 29
- 29
- 29
- Solving Oscillation Problem in Post-Training Quantization Through a Theoretical Perspective28
This paper first formulate the oscillation in PTQ and prove the problem is caused by the difference in module capacity, and then tries to solve this problem by introducing a principled and generalized frame-work theoretically.
- Fast and Accurate Quantized Camera Scene Detection on Smartphones, Mobile AI 2021 Challenge: Report28
The first Mobile AI challenge is introduced, where the target is to develop quantized deep learning-based camera scene classification solutions that can demonstrate a real-time performance on smartphones and IoT platforms.
- Prompt-A-Video: Prompt your Video Diffusion Model via Preference-Aligned LLM25
This work introduces an LLMbased prompt adaptation framework, termed as Prompt-A-Video, which excels in crafting Video-Centric, LaborFree and Preference-Aligned prompts tailored to specific video diffusion model, and validate the effectiveness of Prompt-A-Video across diverse generation models.
- Training-free Diffusion Acceleration with Bottleneck Sampling25
Bottleneck Sampling is introduced, a training-free framework that leverages low-resolution priors to reduce computational overhead while preserving output fidelity while maintaining output quality comparable to the standard full-resolution sampling process across multiple evaluation metrics.
- Progressive Automatic Design of Search Space for One-Shot Neural Architecture Search24
This paper formulate a progressive search strategy based on operation pruning and build a layer-wise operation search space, named PAD-NAS, which can automatically design the operations for each layer and achieve a trade-off between search space quality and model diversity.
- ID-Aligner: Enhancing Identity-Preserving Text-to-Image Generation with Reward Feedback Learning23
The ID-Aligner framework, a general feedback learning framework to enhance ID-T2I performance, introduces identity consistency reward fine-tuning to utilize the feedback from face detection and recognition models to improve generated identity preservation and proposes identity aesthetic reward fine-tuning.
- 21
- MoCoViT: Mobile Convolutional Vision Transformer21
Comprehensive experiments verify that the proposed MoCoViT family outperform state-of-the-art portable CNNs and transformer neural architectures on various vision tasks.
- 21
- Design of a Very Compact CNN Classifier for Online Handwritten Chinese Character Recognition Using DropWeight and Global Pooling21
It is revealed that the novel technique called DropWeight for pruning redundant connections in the CNN architecture not only treats streamlined architectures such as AlexNet and VGGNet well but exhibits remarkable performance for deep residual network and inception network, and is demonstrated that global pooling is a better choice for building very compact online HCCR systems.
- Quantitative analysis for capabilities of vehicular fog computing19
This paper is the first to propose vehicular fog computing by studying its capabilities using realistic data acquired from tens of thousands of taxis in Beijing, China, and quantitatively reveals its capabilities at different scales.
- Dense2MoE: Restructuring Diffusion Transformer to MoE for Efficient Text-to-Image Generation16
This work pioneers the transformation of a dense DiT into a Mixture of Experts (MoE) for structured sparsification, reducing the number of activated parameters while preserving model capacity and establishing a new paradigm for efficient text-to-image generation.
- 15
- 15
- RayFlow: Instance-Aware Diffusion Acceleration via Adaptive Flow Trajectories14
This work proposes RayFlow, a novel diffusion framework that addresses limitations of existing acceleration methods, and introduces Time Sampler, an importance sampling technique to enhance training efficiency by focusing on crucial timesteps.
- 14
- 13
- 13
- IDA-VLM: Towards Movie Understanding via ID-Aware Large Vision-Language Model11
The research introduces a novel benchmark MM-ID, to examine LVLMs on instance IDs memory and recognition across four dimensions: matching, location, question-answering, and captioning, and highlights the limitations of existing LVLMs in recognizing and associating instance identities with ID reference.
- 11
- ResAdapter: Domain Consistent Resolution Adapter for Diffusion Models9
The resolution adapter, a domain-consistent adapter designed for diffusion models to generate images with unrestricted resolutions and aspect ratios, which is compatible with other modules for image generation across a broad range of resolutions.
- UniFL: Improve Latent Diffusion Model via Unified Feedback Learning9
UniFL is presented, a unified framework that leverages feedback learning to enhance diffusion models comprehensively and stands out as a universal, effective, and generalizable solution applicable to various diffusion models, such as SD1.5 and SDXL.
- 9
- Accelerating Content Delivery via Efficient Resource Allocation for Network Coding Aided D2D Communications9
This paper investigates cooperative relaying scheme (RNCC) for cooperative D2D communications, which is aided by the random linear network coding, and proposes a distributed QoS-aware greedy algorithm, which yields close to optimal solution.
- 8
- 8
- 8
- 7
- DLIP: Distilling Language-Image Pre-training7
DLIP, a simple yet efficient Distilling Language-Image Pre-training framework, is presented, through which it is investigated how to distill a light VLP model, and experimental results reveal that DLIP can achieve a state-of-the-art accuracy/efficiency trade-off across diverse cross-modal tasks.
- UGC: Unified GAN Compression for Efficient Image-to-Image Translation6
This work proposes a new learning paradigm, Unified GAN Compression (UGC), with a unified optimization objective to seamlessly prompt the synergy of model-efficient and label-efficient learning.
- Jamcloud: Turning Traffic Jams Into Computation Opportunities–Whose Time Has Come6
This study reveals the microscopic computation capability of local cloudlets as well as the overall and asymptotic performance of the JamCloud, which provides foundational understanding to design, such systems in practice.
- 5
- 5
- 5
- 4
- 4
- 4
- One-step construction of cubic-like NiS2@MoS2 nanocrystals for improved electrocatalytic performance4
- 4
- 4
- 3
- TreeReward: Improve Diffusion Model via Tree-Structured Feedback Learning3
This work introduces a tree-structure reward model to exploit the fine-grained feedback data efficiently and provide tailored optimization during feedback learning and validate the feedback learning performance of the method across different fine-grained dimensions and various downstream tasks.
- 3
- 3
- Multi-Objective Evolutionary for Object Detection Mobile Architectures Search2
A backbone network architecture search algorithm which is a kind of evolutionary optimized method based on non-dominated sorting for NAS scenarios which can quickly search the backbone network architecture within certain computational complexity constraints and better solves the problem of suboptimal linear combination accuracy and computational cost.
- Parallel Pre-trained Transformers (PPT) for Synthetic Data-based Instance Segmentation2
This paper proposes a Parallel Pre-trained Transformers (PPT) framework to accomplish the synthetic data-based Instance Segmentation task, and leverages the off-the-shelf pre-trained vision Transformers to alleviate the gap between natural and synthetic data.
- 2
- LABridge: Text–Image Latent Alignment Framework via Mean-Conditioned OU Process1
Latent Alignment Framework (LABridge), a novel Text–Image Latent Alignment Framework via an Ornstein–Uhlenbeck (OU) Process, which explicitly preserves and aligns textual and visual semantics in an aligned latent space, paves the way for more efficient, semantically consistent, and high-fidelity text to image generation.
- 1
- 1
- 1
- 1
- 1
- –
- Theoretical research of Tm-doped bismuth silicate crystal’s optical properties from first principles–
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-15. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.