7 citations · 12 across the 10 of their papers we have counts for
Showing 2026Show all
2 papers · 1 filter
cs.LG2026
Forking: Sudden Overfitting Under Replay
Shanbin Yu, Shaoyang Guo, Haoran Zhao +2
This paper studies forking, a generalization failure discovered in NanoGPT autoresearch. Under data replay, models with an over-encoding n-gram memory branch show a sharp separatio…
cs.AI2026
Why Do Conventional World Models Fail to Learn Cellular Automata?
Shaoyang Guo, Ziming Liu
Although conventional world models - auto-regressive or diffusion models based on transformers or convolutional networks - may learn surface statistics of world dynamics, can they…