2 papers
cs.LG2026
OrthoFormer: Instrumental Variable Estimation in Transformer Hidden States via Neural Control Functions
Charles Luo
Transformer architectures excel at sequential modeling yet remain fundamentally limited by correlational learning - they capture spurious associations induced by latent confounders…
cs.LG2024
Has LLM Reached the Scaling Ceiling Yet? Unified Insights into LLM Regularities and Constraints
Charles Luo
Large Language Models (LLMs) have demonstrated remarkable capabilities, yet their scalability raises a critical question: Have we reached the scaling ceiling? This paper addresses…