Xiaocheng Feng

Xiaocheng Feng is a prominent researcher in computational linguistics and large language models (LLMs), with extensive contributions to multilingual systems, vision-language integration, and knowledge alignment. His work spans from 2016 to 2025, focusing on advanced topics like

• Mitigating hallucinations in multimodal models
• Efficient knowledge editing across multiple models
• Length-controlled text generation for black-box LLMs
• Improving contextual faithfulness via retrieval heads
• Cross-lingual connection mechanisms for fine-tuning

His recent publications address critical challenges in LLMs, including knowledge misalignment, entity-level unlearning, and Pareto optimization in multilingual translation. He has explored probability density estimation for text control, adaptive context modeling in visual storytelling, and cross-lingual annotation projection for low-resource languages. Collaborations with researchers like Bing Qin, Lei Huang, and Weitao Ma highlight his central role in advancing NLP methodologies.