AI.info

Zhaoyang Li

Explore Zhaoyang Li on AI.info.

• All Changes May Have Invariant Principles: Improving Ever-Shifting Harmful Meme Detection via Design Concept Reproduction
• Know Thy Enemy: Securing LLMs Against Prompt Injection via Diverse Data Synthesis and Instruction-Level Chain-of-Thought Learning
• Towards Robust Pseudo-Label Learning in Semantic Segmentation: An Encoding Perspective
• Balanced Learning for Domain Adaptive Semantic Segmentation
• ACT as Human: Multimodal Large Language Model Data Annotation with Critical Thinking
• 3rd Place Solution to ICCV LargeFineFoodAI Retrieval