Haoru Tan

Scaf-GRPO: Scaffolded Group Relative Policy Optimization for Enhancing LLM Reasoning

Fast Data Mixture Optimization via Gradient Descent

DreamOmni2: Multimodal Instruction-based Editing and Generation

SmartSwitch: Advancing LLM Reasoning by Overcoming Underthinking via Promoting Deeper Thought Exploration

Many Can Beat One: MoE-Linear Attention for Full MLP Image Generation

Multi-Scale Group Relative Policy Optimization for Large Language Models

Understanding Data Influence in Reinforcement Finetuning

Mixture Compressor for Mixture-of-Experts LLMs Gains More

DLoFT: Gradient-Decoupled Fine-Tuning for Generalizable Long Chain-of-Thought Reasoning

Data Pruning by Information Maximization

Diff-In: Data Influence Estimation with Differential Approximation

CR2PQ: Continuous Relative Rotary Positional Query for Dense Visual Representation Learning

Rethinking Out-of-Distribution Detection in Vision Foundation Models

Learn to Synthesize Compact Datasets by Matching Effects