AI研究論文每日精選

MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models

EMOVA: Empowering Language Models to See, Hear and Speak with Vivid Emotions

LLaVA-3D: A Simple yet Effective Pathway to Empowering LMMs with 3D-awareness

Lotus: Diffusion-based Visual Foundation Model for High-quality Dense Prediction

Instruction Following without Instruction Tuning

Discovering the Gems in Early Layers: Accelerating Long-Context LLMs with 1000x Input Token Reduction

Pixel-Space Post-Training of Latent Diffusion Models

The Imperative of Conversation Analysis in the Era of LLMs: A Survey of Tasks, Techniques, and Trends

Disco4D: Disentangled 4D Human Generation and Animation from a Single Image

Reducing the Footprint of Multi-Vector Retrieval with Minimal Performance Impact via Token Pooling

Robot See Robot Do: Imitating Articulated Object Manipulation with Monocular 4D Reconstruction

Enhancing Structured-Data Retrieval with GraphRAG: Soccer Data Case Study