Shan Yu

2 papers indexed

arxivcs.LGcs.AI2026-07-15

Closed-Loop Knowledge Dynamics: An Operational Framework for Saturation and Escape

Feedback-driven loops support iterative improvement in large language models, reinforcement learning, and autonomous discovery, yet their gains often diminish under repeated internal feedback. We study why closed-loop knowledge systems saturate and what external information can m…

View free PDFSource page

arxivcs.AI2026-06-29

BayesEvolve: Explicit Belief States for Autonomous Scientific Discovery

Xuening Wu, Shan Yu, Qianya Xu, Shenqin Yin

Autonomous scientific discovery systems increasingly use large language models (LLMs) to propose new hypotheses, but many such systems condition primarily on experimental memory: archives of high-scoring candidates or heuristic summaries of recent trials. We argue that discovery…

View free PDFSource page