Reading Log
A reading log of the AI papers, books, and resources that shaped how I build and research, each with a short note on what it introduced.
Paper / Transformers / Read
Vaswani et al. (2017)
Introduces the Transformer: self-attention without recurrence or convolution. The backbone of nearly every modern LLM.
Paper / Pretraining / Read
Devlin et al. (2018)
Masked-language-model pretraining for bidirectional context. Reframed NLP around pretrain-then-finetune.
Paper / Scaling / Read
Brown et al. (2020)
Showed that scale alone unlocks in-context / few-shot learning without gradient updates at inference.
Paper / RAG / Reference
Lewis et al. (2020)
The original RAG formulation: combine a parametric LLM with a non-parametric retrieval index. Directly relevant to the chatbots I ship.
Paper / Scaling / Read
Hoffmann et al. (2022)
Corrected the scaling laws: most large models were undertrained on data relative to their parameter count.
Paper / Fine-tuning / Reading
Hu et al. (2021)
Freezes pretrained weights and injects trainable low-rank matrices, so you can fine-tune huge models on modest hardware.
Paper / GNNs / Reference
Wu et al. (2019)
Map of the GNN landscape, and background for my UNSAT-core detection research on bipartite clause graphs.
Book / Foundations / Reference
Goodfellow, Bengio, Courville (2016)
The standard reference for the math and intuition behind deep nets: optimization, regularization, representation.