LatentUM: Unleashing the Potential of Interleaved Cross-Modal Reasoning via a Latent-Space Unified Model
cs.CV, cs.LG
Submitted: 2026-04-02
Updated: 2026-11-15
Terminology
Sources
- Cosmos World Foundation Model Platform for Physical AI
- Ming-Omni: A Unified Multimodal Model for Perception and Generation
- Qwen3-VL Technical Report
- Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
- BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset
- Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling
- Diffusion Autoencoders are Scalable Image Tokenizers
- ANOLE: An Open, Autoregressive, Native Large Multimodal Models for Interleaved Image-Text Generation
- Emu3.5: Native Multimodal Models are World Learners
- Emerging Properties in Unified Multimodal Pretraining
- From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step
- VQRAE: Representation Quantization Autoencoders for Multimodal Understanding, Generation and Reconstruction
- The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding
- MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models
- X-Omni: Reinforcement Learning Makes Discrete Autoregressive Image Generative Models Great Again
- ThinkMorph: Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoning
- World Models
- Vision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations
- UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision
- Uni-X: Mitigating Modality Conflict with a Two-End-Separated Architecture for Unified Multimodal Models
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs