VideoMLA introduces a novel Multi-Head Latent Attention (MLA) mechanism that replaces per-head KV caches with a shared low-rank content latent, significantly reducing memory and improving throughput f…

The paper introduces X-Stream, a new benchmark for multi-stream video understanding, and finds that current state-of-the-art MLLMs perform poorly when required to process multiple concurrent video str…

TunerDiT introduces a training-free progressive steering method to enhance multi-event video generation using Diffusion Transformers, achieving state-of-the-art performance by explicitly managing even…

Real2SAM2Real introduces a framework that uses explicit 3D caches, derived from 3D lifting models, to provide robust geometric guidance to Video Diffusion Models, significantly improving spatiotempora…

The paper proposes a novel render-free framework that conditions video diffusion models directly on compressed 3D human mesh tokens, enabling robust 3D-aware human motion control without relying on re…

The paper systematically characterizes column-level activation sparsity across various diffusion model architectures, demonstrating that element-level sparsity metrics significantly overestimate the a…

BlazeEdit is a highly efficient, generalist image-to-image diffusion model designed for on-device deployment, consolidating multiple editing tasks into a compact 195M parameter model that runs quickly…

The paper proposes a compression pipeline combining few-step distillation and low-bit quantization to significantly reduce the deployment cost and parameter footprint of large dual-expert video diffus…

Lumos-Nexus is a training-efficient framework that enhances video generation quality by progressively bridging generation from a lightweight model to a high-fidelity generator in a shared latent space…

LongLive-RAG proposes a novel Retrieval-Augmented Generation (RAG) framework to stabilize and improve the quality of long-horizon video generation by treating the entire generated history as a searcha…

The paper proposes a novel cross-axis feature fusion architecture and an auxiliary joint-difference prediction task to significantly improve text-based 3D human motion editing by better understanding…

The paper introduces NaRA, a noise-aware LoRA technique that dynamically adapts fine-tuning parameters based on the noise level during diffusion, significantly improving the performance of Diffusion L…

The paper proposes VRPO, a reinforcement learning-based optimization strategy that replaces static alignment losses in diffusion models, significantly improving both convergence and image fidelity.

Chatterbox-Flash introduces a prior-calibrated block diffusion model for zero-shot TTS that achieves high-fidelity, streaming synthesis with significantly lower computational overhead than existing me…