mHC: Manifold-Constrained Hyper-Connections
Paper • 2512.24880 • Published • 318
Read 2026
-
-
Fantastic Reasoning Behaviors and Where to Find Them: Unsupervised Discovery of the Reasoning Process
Paper • 2512.23988 • Published • 19 -
SpaceTimePilot: Generative Rendering of Dynamic Scenes Across Space and Time
Paper • 2512.25075 • Published • 15 -
Guiding a Diffusion Transformer with the Internal Dynamics of Itself
Paper • 2512.24176 • Published • 8 -
DiffThinker: Towards Generative Multimodal Reasoning with Diffusion Models
Paper • 2512.24165 • Published • 52 -
AdaGaR: Adaptive Gabor Representation for Dynamic Scene Reconstruction
Paper • 2601.00796 • Published • 32 -
Taming Preference Mode Collapse via Directional Decoupling Alignment in Diffusion Reinforcement Learning
Paper • 2512.24146 • Published • 14 -
Deep Delta Learning
Paper • 2601.00417 • Published • 34 -
LTX-2: Efficient Joint Audio-Visual Foundation Model
Paper • 2601.03233 • Published • 171 -
SOP: A Scalable Online Post-Training System for Vision-Language-Action Models
Paper • 2601.03044 • Published • 28 -
GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization
Paper • 2601.05242 • Published • 229 -
The Illusion of Specialization: Unveiling the Domain-Invariant "Standing Committee" in Mixture-of-Experts Models
Paper • 2601.03425 • Published • 16 -
RelayLLM: Efficient Reasoning via Collaborative Decoding
Paper • 2601.05167 • Published • 31 -
AgentOCR: Reimagining Agent History via Optical Self-Compression
Paper • 2601.04786 • Published • 30 -
Over-Searching in Search-Augmented Large Language Models
Paper • 2601.05503 • Published • 7 -
Lost in the Noise: How Reasoning Models Fail with Contextual Distractors
Paper • 2601.07226 • Published • 33 -
Beyond Hard Masks: Progressive Token Evolution for Diffusion Language Models
Paper • 2601.07351 • Published • 26 -
Dr. Zero: Self-Evolving Search Agents without Training Data
Paper • 2601.07055 • Published • 22 -
User-Oriented Multi-Turn Dialogue Generation with Tool Use at scale
Paper • 2601.08225 • Published • 53 -
The Confidence Dichotomy: Analyzing and Mitigating Miscalibration in Tool-Use Agents
Paper • 2601.07264 • Published • 24 -
SnapGen++: Unleashing Diffusion Transformers for Efficient High-Fidelity Image Generation on Edge Devices
Paper • 2601.08303 • Published • 19 -
Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planning
Paper • 2601.09708 • Published • 54 -
Flow Equivariant World Models: Memory for Partially Observed Dynamic Environments
Paper • 2601.01075 • Published • 6 -
The AI Hippocampus: How Far are We From Human Memory?
Paper • 2601.09113 • Published • 6 -
Rewarding the Rare: Uniqueness-Aware RL for Creative Problem Solving in LLMs
Paper • 2601.08763 • Published • 149 -
Alterbute: Editing Intrinsic Attributes of Objects in Images
Paper • 2601.10714 • Published • 31 -
Transition Matching Distillation for Fast Video Generation
Paper • 2601.09881 • Published • 33 -
Unlocking Implicit Experience: Synthesizing Tool-Use Trajectories from Text
Paper • 2601.10355 • Published • 39 -
Language of Thought Shapes Output Diversity in Large Language Models
Paper • 2601.11227 • Published • 9 -
More Images, More Problems? A Controlled Analysis of VLM Failure Modes
Paper • 2601.07812 • Published • 6 -
Toward Efficient Agents: Memory, Tool learning, and Planning
Paper • 2601.14192 • Published • 57 -
Agentic Reasoning for Large Language Models
Paper • 2601.12538 • Published • 202 -
Scaling Text-to-Image Diffusion Transformers with Representation Autoencoders
Paper • 2601.16208 • Published • 55 -
PROGRESSLM: Towards Progress Reasoning in Vision-Language Models
Paper • 2601.15224 • Published • 12 -
360Anything: Geometry-Free Lifting of Images and Videos to 360°
Paper • 2601.16192 • Published • 9 -
Agentic Uncertainty Quantification
Paper • 2601.15703 • Published • 9 -
MeepleLM: A Virtual Playtester Simulating Diverse Subjective Experiences
Paper • 2601.07251 • Published • 11 -
C-RADIOv4 (Tech Report)
Paper • 2601.17237 • Published • 10 -
Agentic Very Long Video Understanding
Paper • 2601.18157 • Published • 19 -
Shaping capabilities with token-level data filtering
Paper • 2601.21571 • Published • 27 -
KromHC: Manifold-Constrained Hyper-Connections with Kronecker-Product Residual Matrices
Paper • 2601.21579 • Published • 6 -
Golden Goose: A Simple Trick to Synthesize Unlimited RLVR Tasks from Unverifiable Internet Text
Paper • 2601.22975 • Published • 109 -
MemOCR: Layout-Aware Visual Memory for Efficient Long-Horizon Reasoning
Paper • 2601.21468 • Published • 25 -
LMK > CLS: Landmark Pooling for Dense Embeddings
Paper • 2601.21525 • Published • 5 -
Diversity-Preserved Distribution Matching Distillation for Fast Visual Synthesis
Paper • 2602.03139 • Published • 45 -
Rethinking the Trust Region in LLM Reinforcement Learning
Paper • 2602.04879 • Published • 37 -
Protein Autoregressive Modeling via Multiscale Structure Generation
Paper • 2602.04883 • Published • 3 -
MSign: An Optimizer Preventing Training Instability in Large Language Models via Stable Rank Restoration
Paper • 2602.01734 • Published • 32 -
Weak-Driven Learning: How Weak Agents make Strong Agents Stronger
Paper • 2602.08222 • Published • 283 -
Towards Agentic Intelligence for Materials Science
Paper • 2602.00169 • Published • 47 -
Reliable and Responsible Foundation Models: A Comprehensive Survey
Paper • 2602.08145 • Published • 8 -
Col-Bandit: Zero-Shot Query-Time Pruning for Late-Interaction Retrieval
Paper • 2602.02827 • Published • 3 -
Stable Velocity: A Variance Perspective on Flow Matching
Paper • 2602.05435 • Published • 3 -
The Pensieve Paradigm: Stateful Language Models Mastering Their Own Context
Paper • 2602.12108 • Published • 13 -
Free(): Learning to Forget in Malloc-Only Reasoning Models
Paper • 2602.08030 • Published • 6 -
When the Prompt Becomes Visual: Vision-Centric Jailbreak Attacks for Large Image Editing Models
Paper • 2602.10179 • Published • 6 -
GeoAgent: Learning to Geolocate Everywhere with Reinforced Geographic Characteristics
Paper • 2602.12617 • Published • 20 -
Experiential Reinforcement Learning
Paper • 2602.13949 • Published • 71 -
SPILLage: Agentic Oversharing on the Web
Paper • 2602.13516 • Published -
Exposing the Systematic Vulnerability of Open-Weight Models to Prefill Attacks
Paper • 2602.14689 • Published • 1 -
On Surprising Effectiveness of Masking Updates in Adaptive Optimizers
Paper • 2602.15322 • Published • 10 -
Visual Persuasion: What Influences Decisions of Vision-Language Models?
Paper • 2602.15278 • Published • 3 -
The Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent Systems
Paper • 2602.15382 • Published • 2 -
Causal-JEPA: Learning World Models through Object-Level Latent Interventions
Paper • 2602.11389 • Published • 7 -
Empty Shelves or Lost Keys? Recall Is the Bottleneck for Parametric Factuality
Paper • 2602.14080 • Published • 20 -
Multi-agent cooperation through in-context co-player inference
Paper • 2602.16301 • Published • 24 -
Reinforced Fast Weights with Next-Sequence Prediction
Paper • 2602.16704 • Published • 13 -
World Action Models are Zero-shot Policies
Paper • 2602.15922 • Published • 14 -
Unified Latents (UL): How to train your latents
Paper • 2602.17270 • Published • 58 -
DDiT: Dynamic Patch Scheduling for Efficient Diffusion Transformers
Paper • 2602.16968 • Published • 12 -
NeST: Neuron Selective Tuning for LLM Safety
Paper • 2602.16835 • Published • 1 -
CrispEdit: Low-Curvature Projections for Scalable Non-Destructive LLM Editing
Paper • 2602.15823 • Published • 3 -
Does Your Reasoning Model Implicitly Know When to Stop Thinking?
Paper • 2602.08354 • Published • 262 -
Spanning the Visual Analogy Space with a Weight Basis of LoRAs
Paper • 2602.15727 • Published • 14 -
VLANeXt: Recipes for Building Strong VLA Models
Paper • 2602.18532 • Published • 52 -
On Data Engineering for Scaling LLM Terminal Capabilities
Paper • 2602.21193 • Published • 99 -
Test-Time Training with KV Binding Is Secretly Linear Attention
Paper • 2602.21204 • Published • 30 -
One-step Language Modeling via Continuous Denoising
Paper • 2602.16813 • Published • 4 -
Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking
Paper • 2602.21196 • Published • 5 -
The Diffusion Duality, Chapter II: Ψ-Samplers and Efficient Curriculum
Paper • 2602.21185 • Published • 3 -
DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference
Paper • 2602.21548 • Published • 46 -
Image Generation with a Sphere Encoder
Paper • 2602.15030 • Published • 15 -
From Statics to Dynamics: Physics-Aware Image Editing with Latent Transition Priors
Paper • 2602.21778 • Published • 14 -
SeaCache: Spectral-Evolution-Aware Cache for Accelerating Diffusion Models
Paper • 2602.18993 • Published • 4 -
Dropping Anchor and Spherical Harmonics for Sparse-view Gaussian Splatting
Paper • 2602.20933 • Published • 4 -
VGG-T^3: Offline Feed-Forward 3D Reconstruction at Scale
Paper • 2602.23361 • Published • 14 -
Causal Motion Diffusion Models for Autoregressive Motion Generation
Paper • 2602.22594 • Published • 7 -
AgentDropoutV2: Optimizing Information Flow in Multi-Agent Systems via Test-Time Rectify-or-Reject Pruning
Paper • 2602.23258 • Published • 28 -
Mode Seeking meets Mean Seeking for Fast Long Video Generation
Paper • 2602.24289 • Published • 41 -
LK Losses: Direct Acceptance Rate Optimization for Speculative Decoding
Paper • 2602.23881 • Published • 18 -
How to Take a Memorable Picture? Empowering Users with Actionable Feedback
Paper • 2602.21877 • Published • 14 -
Spectral Condition for μP under Width-Depth Scaling
Paper • 2603.00541 • Published • 15 -
Beyond Language Modeling: An Exploration of Multimodal Pretraining
Paper • 2603.03276 • Published • 99 -
Spilled Energy in Large Language Models
Paper • 2602.18671 • Published • 11 -
Helios: Real Real-Time Long Video Generation Model
Paper • 2603.04379 • Published • 173 -
RealWonder: Real-Time Physical Action-Conditioned Video Generation
Paper • 2603.05449 • Published • 12 -
KARL: Knowledge Agents via Reinforcement Learning
Paper • 2603.05218 • Published • 6 -
Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling
Paper • 2603.04553 • Published • 3 -
Progressive Residual Warmup for Language Model Pretraining
Paper • 2603.05369 • Published • 36 -
Dynamic Model Routing and Cascading for Efficient LLM Inference: A Survey
Paper • 2603.04445 • Published • 4 -
MM-Zero: Self-Evolving Multi-Model Vision Language Models From Zero Data
Paper • 2603.09206 • Published • 51 -
Lost in Backpropagation: The LM Head is a Gradient Bottleneck
Paper • 2603.10145 • Published • 11 -
Just-in-Time: Training-Free Spatial Acceleration for Diffusion Transformers
Paper • 2603.10744 • Published • 7 -
Hindsight Credit Assignment for Long-Horizon LLM Agents
Paper • 2603.08754 • Published • 5 -
HomeSafe-Bench: Evaluating Vision-Language Models on Unsafe Action Detection for Embodied Agents in Household Scenarios
Paper • 2603.11975 • Published • 11 -
Grounding World Simulation Models in a Real-World Metropolis
Paper • 2603.15583 • Published • 142 -
Attention Residuals
Paper • 2603.15031 • Published • 136 -
WiT: Waypoint Diffusion Transformers via Trajectory Conflict Navigation
Paper • 2603.15132 • Published • 33 -
LoST: Level of Semantics Tokenization for 3D Shapes
Paper • 2603.17995 • Published • 25 -
V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning
Paper • 2603.14482 • Published • 13 -
ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents
Paper • 2603.18815 • Published • 8 -
COT-FM: Cluster-wise Optimal Transport Flow Matching
Paper • 2603.13395 • Published • 1 -
Matryoshka Gaussian Splatting
Paper • 2603.19234 • Published • 6