论文

• Autoregressive Language Models are Secretly Energy-Based Models: Insights into the Lookahead Capabilities of Next-Token Prediction
• Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free
• GenEval 2: Addressing Benchmark Drift in Text-to-Image Evaluation
• Towards Scalable Pre-training of Visual Tokenizers for Generation
• Qwen-Image-Layered: Towards Inherent Editability via Layer Decomposition
• The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding
• Meta-RL Induces Exploration in Language Agents
• Qwen3-VL Technical Report
• DeepStack: Deeply Stacking Visual Tokens is Surprisingly Simple and Effective for LMMs
• Hyper-Connections
• mHC: Manifold-Constrained Hyper-Connections
• Next-Latent Prediction Transformers Learn Compact World Models
• NextFlow: Unified Sequential Modeling Activates Multimodal Understanding and Generation
• Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models
• Gramian Multimodal Representation Learning and Alignment