- llm
- nlp
- reasoning
- rl
- rlhf
- ssm
- theory
- information-theory
- representation-learning
- optimization
- rag
- alignment
- paper-review
- survey
- notes
•
•
•
•
•
•
•
•
•
•
•
•
•
•
-
Graph of Thoughts: Solving Elaborate Problems with LLMs
Generalizing CoT and ToT to arbitrary graph-structured reasoning over LLM thoughts.
-
Sparks of AGI: Early Experiments with GPT-4
Notes on the Microsoft paper exploring GPT-4's broad capabilities across reasoning, coding, math, and multimodal tasks.
-
Neural Tangent Kernel: Infinite-Width Networks as Kernel Methods
How wide neural networks behave as kernel regression — the NTK framework.
-
Mini-batch Optimization of Contrastive Loss
Sample complexity and optimization dynamics of contrastive loss under mini-batch training.
-
Frequency Effects on Syntactic Rule Learning in Transformers
ACL 2021 — how token frequency in training data shapes syntactic generalization in Transformers.