Ren Chen

The paper proposes DAG-MoE, a novel sparse Mixture-of-Experts framework that replaces standard weighted-sum aggregation with structural aggregation to enhance model performance and enable multi-step reasoning.

The paper introduces Grounded Agentic Interaction Synthesis (GAIS), a framework that generates high-quality, diverse, and complex agentic training data by anchoring tasks to real-world protocols, significantly improving base model performance.

The paper introduces State-Adaptive Prompt Optimization (SAPO), a novel training strategy that treats prompts as dynamic variables to achieve robust fine-tuning, significantly mitigating catastrophic forgetting and improving generalization in LLMs.