READING LIBRARY

Find a paper by engineering question

Filter by field, reading depth, or the paper’s publication year.

Showing 24 of 106 notes

EXPLORE BY RESEARCH TOPIC

Explore the library by research topic

The latest notes are listed above. These topic paths collect every relevant deep dive, even when a paper belongs to more than one engineering question.

4 notes

Sequence Modeling Foundations

Build a foundation for reading sequence-transduction papers through classic encoder–decoder and self-attention architectures.

Question:Where does a sequence model change its control point, and which evidence still transfers to today’s systems?

  1. Transformer: Drop Recurrence for Self-Attention, but WMT 2017 BLEU Does Not Represent Later LLMs
  2. InstructGPT: Align Instructions with Human Feedback, but 2022 Preference Win Rates Do Not Represent Later ChatGPT
View this topic
2 notes

Generative Media & Interactive Systems

Study state, control scope and continuity in generated artifacts—from symbolic score editing to interactive world generation—and the limits of their evaluation evidence.

Question:How can a generative system edit the current artifact, preserve prior results, and apply controls only within their intended scope?

  1. ABC-LLM: How Can AI Edit a Passage and Preserve the Rest of the Score?
  2. Hy-Nomas: remembering scenes while letting actions end
View this topic

START WITH A PATH

Choose the question you want to learn

Start with three notes in each path. Continue with optional readings when you need more depth.

Build context with How to Read AI Agent Papers: From CoT and WebGPT to ReAct and How to Read RAG Papers: From Dense Retrieval (DPR) to Lewis RAG.

Intro → Intermediate 7 reads

Build the foundations first

Start with AlexNet architecture and training evidence, then ResNet residual shortcuts for trainable depth, YOLO unified real-time detection as single-pass regression, Transformer sequence transduction with self-attention, InstructGPT instruction alignment via SFT, a reward model, and PPO, then Speculative Decoding lossless inference via draft-and-verify parallel decoding, while separating methods, experiments, and historical constraints.

Start with 3 notes · about 47 min

  1. AlexNet Part 1: Reading the Evidence Behind an ImageNet Turning Point
  2. AlexNet Part 2: Turning the Training Recipe into Testable Design Choices
  3. ResNet: Residuals Make Depth Trainable, but ImageNet 2015 Is Not a Ready-Made Detection or ViT Contract

+ 4 optional readings

Start this path
Intermediate → Advanced 31 reads

Retrieval, memory, and production RAG

Start from 2020 retrieval-augmented pre-training (REALM), then the cheaper dense retriever (DPR), the RAG ancestor and Self-RAG when-to-retrieve, then multimodal parsing, tool retrieval, memory, GraphRAG, scaling, and runtime controls.

Start with 3 notes · about 57 min

  1. REALM: Wire Retrieval into LM Pre-Training, but Do Not Treat Joint Training as a Ready-Made RAG Stack
  2. DPR: Turn Open-Domain QA into Dense Passage Retrieval, but Do Not Treat the Dual Encoder as Production RAG
  3. RAG: Attach Retrieval to Generation, but Do Not Treat 2020 RAG as a Production RAG Platform

+ 28 optional readings

Start this path
Advanced 66 reads

Agent runtime, safety, and evaluation

Start with canonical 2017–2023 agent method ancestors (CoT → WebGPT → ReAct → Toolformer → Gorilla → IPI → SWE-bench → Reflexion → MemGPT → Generative Agents), then later tool/eval, runtime, safety, platform, and multimodal end-to-end work evaluation.

Start with 3 notes · about 60 min

  1. CoT: Make the Model Write the Reasoning, but Do Not Treat It as an Agent That Moves
  2. WebGPT: Let the Model Browse for Answers, but Do Not Treat It as a Reasoning Agent Loop
  3. ReAct: Interleave Thought and Action, but Do Not Treat a Few-Shot Loop as an Agent Runtime

+ 63 optional readings

Start this path
Intro: problem and intuition Intermediate: method and experiments Advanced: evidence boundaries and adoption

RESEARCH EXCHANGE

Turn a paper into an engineering discussion

Speaking & contact