Blog
告别冗长 PDF，一文读懂最新顶会与核心期刊的创新价值。
Normalizing Trajectory Models
SOD: Step-wise On-policy Distillation for Small Language Model Agents
TextLDM: Language Modeling with Continuous Latent Diffusion
Memory-Efficient Looped Transformer: Decoupling Compute from Memory in Looped Language Models
SimCT: Recovering Lost Supervision for Cross-Tokenizer On-Policy Distillation
Drifting Field Policy: A One-Step Generative Policy via Wasserstein Gradient Flow
Anisotropic Modality Align
Decoherence without the state: A causal quantum Darwinist approach
Fast Byte Latent Transformer
InterLV-Search: Benchmarking Interleaved Multimodal Agentic Search
Towards Differentially Private Reinforcement Learning with General Function Approximation
Scalable Active Metamaterials for Shape-Morphing
Post-training makes large language models less human-like
The Structural Origin of Attention Sink: Variance Discrepancy, Super Neurons, and Dimension Disparity
123D: Unifying Multi-Modal Autonomous Driving Data at Scale
NoiseGate: Learning Per-Latent Timestep Schedules as Information Gating in World Action Models
RELO: Reinforcement Learning to Localize for Visual Object Tracking
One Token Per Frame: Reconsidering Visual Bandwidth in World Models for VLA Policy
Active Embodiment Identification with Reinforcement Learning for Legged Robots
Coupling Models for One-Step Discrete Generation
SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild
STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation
Beyond Pairs: Your Language Model is Secretly Optimizing a Preference Graph
HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents
Machine Learning Techniques for Astrophysics and Cosmology: Photometric Redshifts
Topological Characterization of Discrete-Time Classical Stochastic Processes: Dual Role of Point-Gap Topology
Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Models
FAB: A First-Order AB-based Gradient Algorithm for Distributed Bilevel Optimization over Time-Varying Directed Graphs
Prune-OPD: Efficient and Reliable On-Policy Distillation for Long-Horizon Reasoning
Genus drop involving non-hyperelliptic curves in Feynman integrals