学术相关
CS 8803 LLM - Large Language Models (Georgia Tech)
- Pretraining
- Embeddings
- MoE
- Reasoning
- Agent Harness
- Long-Context
- Attention
- Reinforcement Learning
- Self-Play RL
- Test-Time Scaling
- Mode Collapse
- Linear Transformer
- Diffusion LM
- Safety
- Mechanistic Interpretability
- Calibration
- Scaling Law
Model discovery agent: LLM-assisted bayesian experiment design for data-efficient discovery of mechanistic world models1
让 LLM 负责”提出假设”，让经典贝叶斯统计负责”检验假设、设计实验”，两者组成一个闭环 agent，用极少的主动实验从黑箱系统中发现真正的机理（因果）模型。
Consumer inferences from product rankings: the role of beliefs in search behavior2
业界动态
随机森林之外丨Looping因子挖掘：Agent量化的迭代尽头，真实Alpha有多少？
Goodhart‘s Law: 当一个指标变成考核目标，它就不再是一个好的度量标准。
Loop Engineering的目标设定要注意边界问题，避免过拟合
-
Murphy, K. (2026). Model discovery agent: LLM-assisted bayesian experiment design for data-efficient discovery of mechanistic world models (Version 3). arXiv. https://doi.org/10.48550/ARXIV.2608.09696 ↩
-
Fong, J., Natan, O. R., & Pantle, R. (2026). Consumer inferences from product rankings: The role of beliefs in search behavior. Management Science. https://doi.org/10.1287/mnsc.2025.00761 ↩