Pengxiang Li
- Adaptive Classifier-Free Guidance via Dynamic Low-Confidence Masking
- From Objects to Anywhere: A Holistic Benchmark for Multi-level Visual Grounding in 3D Scenes
- GPAS: Accelerating Convergence of LLM Pretraining via Gradient-Preserving Activation Scaling
- Iterative Tool Usage Exploration for Multimodal Agents via Step-wise Preference Tuning
- The Curse of Depth in Large Language Models