Co-authored publications
Publications and ongoing work on evaluation, agents, reasoning, and alignment.
Xiangqi Wang in bold
-
DyFlow: Dynamic Workflow Framework for Agentic Reasoning
Dynamic workflow construction for tool-using and multi-agent reasoning, supported by theoretical analysis.
-
CLIPErase: Efficient Unlearning of Visual-Textual Associations in CLIP
Contributed to implementation and research visualization.
-
On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective
Contributed to the assessment framework, methodology, and large-scale experiments.
-
Think it Image by Image: Multi-Image Moral Reasoning of Large Vision-Language Models
Contributed to manuscript development and experimental analysis.
-
Exposing and Patching the Flaws of Large Language Models in Social Character Simulation
Related preprint: Social Science Meets LLMs: How Reliable Are Large Language Models in Social Simulations? (2024; arXiv:2410.23426). Contributed simulation scenarios, reliability metrics, and analysis.
-
ValueLence: A Dashboard for In-Depth Value Probing of LLMs
Contributed to the evaluation and analysis pipeline.
No papers match that search. Try a different topic or .