Jun Zhu

DiffusionNFT: Online Diffusion Reinforcement with Forward Process

Exploratory Diffusion Model for Unsupervised Reinforcement Learning

Part-X-MLLM: Part-aware 3D Multimodal Large Language Model

Diffusion Models as Dataset Distillation Priors

Nano3D: A Training-Free Approach for Efficient 3D Editing Without Masks

Mitigating Object Hallucination in Large Vision-Language Models through Adversarial Contrastive Finetuning

Theoretical Analysis of Relative Errors in Gradient Computations for Adversarial Attacks with CE Loss

Helix: Evolutionary Reinforcement Learning for Open-Ended Scientific Problem Solving

AnyPos: Automated Task-Agnostic Actions for Bimanual Manipulation

UltraViCo: Breaking Extrapolation Limits in Video Diffusion Transformers

Large Scale Diffusion Distillation via Score-Regularized Continuous-Time Consistency

Towards Safe Reasoning in Large Reasoning Models via Corrective Intervention

SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse–Linear Attention

Jailbreaking LLMs' Safeguard with Universal Magic Words for Text Embedding Models

MePo: Meta Post-Refinement for Rehearsal-Free General Continual Learning

Vidarc: Low Latency Embodied Video Diffusion Model with Closed-loop Control

Unveiling the Basin-Like Loss Landscape in Large Language Models

NFT: Bridging Supervised Learning and Reinforcement Learning in Math Reasoning

VoiceBridge: Designing Latent Bridge Models for General Speech Restoration at Scale

Efficient Hyperparameter Tuning via Trajectory Invariance Principle

Vidar: Embodied Video Diffusion Model for Generalist Manipulation

Multimodal Physical Adversarial Clothing Evades Visible-Thermal Detectors with Non-Overlapping RGB-T Pattern

Exploring Aleatoric Uncertainty in Object Detection via Vision Foundation Models

Towards the Worst-case Robustness of Large Language Models

LangSAM: Language-Guided Expert Routing on SAM2 for Dense Scene Understanding

AudioMoG: Guiding Audio Generation with Mixture-of-Guidance

Stabilizing Gradient Descent via Second-Order Control-Theoretic Dynamics

STAIR: Improving Safety Alignment with Introspective Reasoning

SageAttention2: Efficient Attention with Thorough Outlier Smoothing and Per-thread INT4 Quantization

On the Optimization and Generalization of Two-layer Transformers with Sign Gradient Descent

Audio Super-Resolution with Latent Bridge Models

RIFLEx: A Free Lunch for Length Extrapolation in Video Diffusion Transformers

Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator

Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment

SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Acceleration

Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Sampling

RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation

ShapeLLM-Omni: A Native Multimodal LLM for 3D Generation and Understanding

Scaling Diffusion Transformers Efficiently via $\mu$P

SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training

Robust Representation Consistency Model via Contrastive Denoising

PivotMesh: Generic 3D Mesh Generation via Pivot Vertices Guidance

FrameBridge: Improving Image-to-Video Generation with Bridge Models

Visual Generation Without Guidance

SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference

ReMoE: Fully Differentiable Mixture-of-Experts with ReLU Routing

A Regularized Newton Method for Nonconvex Optimization with Global and Local Complexity Guarantees

Oscillation-Reduced MXFP4 Training for Vision Transformers

Diffusion Bridge Implicit Models

Elucidating the Preconditioning in Consistency Distillation

ManiBox: Enhancing Spatial Grasping Generalization via Scalable Simulation Data Generation

Zero-shot Quantization for Object Detection

When Bigger is Better: Revisiting Large-Batch Optimization in Language Model Pretraining

FrameBridge: Improving Image-to-Video Generation with Bridge Models

SparseDM: Toward Sparse Efficient Diffusion Models

LUNCH: Adaptive Balancing of Continual Learning via Hyperparameter Uncertainty