maddingface's picture
Update logbook: Repro: Syntax vs. Semantics: How Transformers Learn Deep Dependencies
c92cc38 verified
|
Raw History Blame Contribute Delete
385 Bytes
# Repro: Syntax vs. Semantics: How Transformers Learn Deep Dependencies
## Pages
| Page |
| --- |
| [Claim 1: Gradient Starvation and Phase Transition](#/claim-1-gradient-starvation-and-phase-transition) |
| [Claim 2: Chain-of-Thought and Topology-Aligned Contrastive Objective](#/claim-2-chain-of-thought-and-topology-aligned-contrastive-objective) |
| [Conclusion](#/conclusion) |