Yu Cheng

AdaReasoner: Dynamic Tool Orchestration for Iterative Visual Reasoning

Spotlight on Token Perception for Multimodal Reinforcement Learning

ExGRPO: Learning to Reason from Experience

MoM: Linear Sequence Modeling with Mixture-of-Memories

Interleaving Reasoning for Better Text-to-Image Generation

FrameThinker: Learning to Think with Long Videos via Multi-Turn Frame Spotlighting

DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models

3D-CoS: A New 3D Reconstruction Paradigm Based on VLM Code Synthesis

Conditional Advantage Estimation for Reinforcement Learning in Large Reasoning Models

Reasoning over Boundaries: Enhancing Specification Alignment via Test-time Delibration

Invariant Predict-and-Combinatorial Optimization under Distribution Shifts

ThinkMorph: Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoning

Diversity-Incentivized Exploration for Versatile Reasoning

From Broad Exploration to Stable Synthesis: Entropy-Guided Optimization for Autoregressive Image Generation

ExpertWeaver: Unlocking the Inherent MoE in Dense LLMs with GLU Activation Patterns

FullFront: Benchmarking MLLMs Across the Full Front-End Engineering Workflow

Flash-DMD: Unifying Distillation and Refinement for High-Fidelity Few-Step Image Generation

Revisual-R1: Advancing Multimodal Reasoning From Optimized Cold Start to Staged Reinforcement Learning

Towards an AI Musician: Synthesizing Sheet Music Problems for Musical Reasoning

Process Reinforcement through Implicit Rewards

Scaling Reasoning, Losing Control: Evaluating Instruction Following in Large Reasoning Models

Rethinking Entropy Regularization in Large Reasoning Models

HiPhO: How Far Are (M)LLMs from Humans in the Latest High School Physics Olympiad Benchmark?

Are Unified Vision-Language Models Necessary: Generalization Across Understanding and Generation

Structured-Aware Pre-trained LM Can Improve RNA Structured Prediction

Nirvana: A Specialized Generalist Model With Task-Aware Memory Mechanism

The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models

Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark

Learning to Reason under Off-Policy Guidance

Liger: Linearizing Large Language Models to Gated Recurrent Structures

Occult: Optimizing Collaborative Communications across Experts for Accelerated Parallel MoE Training and Inference

Divide and Conquer: Grounding LLMs as Efficient Decision-Making Agents via Offline Hierarchical Reinforcement Learning

Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation

VideoREPA: Learning Physics for Video Generation through Relational Alignment with Foundation Models

Text-to-Decision Agent: Offline Meta-Reinforcement Learning from Natural Language Supervision

Diving into Self-Evolving Training for Multimodal Reasoning

Make LoRA Great Again: Boosting LoRA with Adaptive Singular Values and Mixture-of-Experts Optimization Alignment

LASP-2: Rethinking Sequence Parallelism for Linear Attention and its Hybrid

Modality-Specialized Synergizers for Interleaved Vision-Language Generalists

RoRA-VLM: Robust Retrieval-Augmented Vision Language Models

Weak to Strong Generalization for Large Language Models with Multi-capabilities

Unveiling the Compositional Ability Gap in Vision-Language Reasoning Model

Scaling Laws for Floating–Point Quantization Training

CLIP-MoE: Towards Building Mixture of Experts for CLIP with Diversified Multiplet Upcycling

Diving into Self-Evolve Training for Multimodal Reasoning

Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation

Test-Time Preference Optimization: On-the-Fly Alignment via Iterative Textual Feedback

Examining Post-Training Quantization for Mixture-of-Experts: A Benchmark

Demystifying the Underappreciated Long-Tail Problems in Large Vision Language Models

Inv-PnCO: Invariant Predict-and-Combinatorial Optimization under Distribution Shifts

Grounded-VideoLLM: Sharpening Fine-grained Temporal Grounding in Video Large Language Models

Strategy-centric Synthesis: Connecting Billions of Image-Text Pairs to High-Quality Visual Instruction Data