Papers, code, notebooks, documentation, tutorials, and the ideas and implementations made by, in the orbit of, and of interest to the group of (semi-)anonymous humans and AI that has been described as the "modern day bell labs".
i'm still unpacking all of my resources and curating! with multiple lines of research happening between a few dozen people flowing in and out and often asynchronous, but continuous context, things will get missed. open a PR. this is a growing repository, not currently inclusive of past, present, and future work. this repo likes:
- research and engineering to solve problems
- in depth resources made with care and love
- a general avoidance of product placements, proprietary tools and services
- things I ultimately find are tasteful additions
- obsessive notion drops and blog posts are high signal
where available, direct references to the authors and creators are provided.
out of scope for this repo are the many excellent resources of the "standard" papers, how to learn ML, etc. I welcome "other topics" PRs to those lists, but I myself do not track that other than the few that I directly send others that more fit this format.
this is not an official work or endorsed in any way. i was looking for a corpus of ideas and resources to separate the signal from noise for others, and this is my attempt at doing so.
CC0 1.0 Universal - this repository provides resources that are under a variety of open licensing. as such, this curation is in the public domain.
- Generative AI Handbook: A Roadmap for Learning Resources - @willccbb a comprehensive starting guide updated to June 2024
- entropix - @_xjdr - what started it all, by swamp ogre himself
- Chain of Thought Empowers Transformers to Solve Inherently Serial Problems - understanding the "how" of chain of thought works and how it can scale
- AlphaGo Zero: Starting from scratch - improving upon AlphaGo without any prior knowledge via reinforcement learning
- MuZero: Mastering Go, chess, shogi, and Atari without rules
- DeepSeek LLM: Scaling Open-Source Language Models with Longtermism
- DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models
- DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- DeepSeek-VL: Towards Real-World Vision-Language Understanding
- DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
- DeepSeek-Prover: Advancing Theorem Proving in LLMs through Large-Scale Synthetic Data
- DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence
- Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models
- DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search
- Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation
- JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation
- DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding
- DeepSeek-V3 Technical Report
still reading background materials but of particular interest - stay tuned
- Efficiently Scaling Transformer Inference - toward maximizing as much of the compute that is available
- Open Problems in Mechanistic Interpretability - an up-to-date state of unresolved issues in mech interp
- a reinforcement learning guide - @naklecha - a lovely breakdown of the fundamentals of reinforcement learning
- Qwen 0.5b on GRPO - @Dorialexander - see the magic of thinking by reinforcement learning with verifiable rewards, all in the space of a free google colab environment
this is my jam, a massive amount coming here shortly
again, this is a curation, so you should cite the resources themselves, but if you really want to cite this, cite it as:
@misc{gerred_awesome_hensen,
author = {Gerred Dillon},
title = {Awesome Hensen},
year = {2025},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/gerred/awesome-hensen}},
}