Song Han

ParoQuant: Pairwise Rotation Quantization for Efficient Reasoning LLM Inference

Locality-aware Parallel Decoding for Efficient Autoregressive Image Generation

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding

LongLive: Real-time Interactive Long Video Generation

SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer

StreamingVLM: Real-Time Understanding for Infinite Video Streams

Fast-dLLM v2: Efficient Block-Diffusion LLM

QeRL: Beyond Efficiency - Quantization-enhanced Reinforcement Learning for LLMs

3D Aware Region Prompted Vision Language Model

Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed

DC-Gen: Post-Training Diffusion Acceleration with Deeply Compressed Latent Space

OmniVinci: Enhancing Architecture and Data for Omni-Modal Understanding LLM

Optimizing Mixture of Block Attention

DC-VideoGen: Efficient Video Generation with Deep Compression Video Autoencoder

FoundationMotion: Auto-labeling and Reasoning about Spatial Movement in Videos

EgoVLA: Learning Vision-Language-Action Models from Egocentric Human Videos

SANA: Efficient High-Resolution Text-to-Image Synthesis with Linear Diffusion Transformers

Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning

SVDQuant: Absorbing Outliers by Low-Rank Component for 4-Bit Diffusion Models

Sparse VideoGen2: Accelerate Video Generation with Sparse Attention via Semantic-Aware Permutation

Radial Attention: $\mathcal O(n \log n)$ Sparse Attention for Long Video Generation

SANA 1.5: Efficient Scaling of Training-Time and Inference-Time Compute in Linear Diffusion Transformer

Jet-Nemotron: Efficient Language Model with Post Neural Architecture Search

Scaling RL to Long Videos

Deep Compression Autoencoder for Efficient High-Resolution Diffusion Models

HART: Efficient Visual Generation with Hybrid Autoregressive Transformer

LongVILA: Scaling Long-Context Visual Language Models for Long Videos

VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation

DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads

XAttention: Block Sparse Attention with Antidiagonal Scoring

Sparse Video-Gen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity

COAT: Compressing Optimizer states and Activations for Memory-Efficient FP8 Training

SparseLoRA: Accelerating LLM Fine-Tuning with Contextual Sparsity

X-VILA: Cross-Modality Alignment for Large Language Models

Wolf: Accurate Video Captioning with a World Summarization Framework

VILA^2: VLM Augmented VLM with Self-Improvement