Active AI Reading

We read AI papers the way researchers do: we rewrite them, criticize them, redraw their figures, rerun their code, trace their ideas and tell the world about them. Everything here is written by the students, in the open.

Browse papers How the course works Write a post
✍️ Writing 🧠 Learning to learn 📊 Visualization 📣 Communicate to the world 🔬 Research skills 💻 Coding 🔭 Paper scouting 😎 Chill

Papers

Why Pretraining Fails to Share Cross-Lingual Knowledge
Adam Gaber, Uriel Dolev, Elisabeth Fittschen, Bobby Cheng,... · 2026
Bilingual LLMs learn the same fact twice, once per language, and the cause is largely tokenization: two copies of English with disjoint token IDs already fail to share knowledge. Mapping Arabic words...
✍️ 2🧠 1📊 2📣 2🔬 3💻 1
📄
Matryoshka Language Model Suites
Nathan Godey, Yoav Artzi · 2026
(To be written by the class after week 2. Edit this file: one or two sentences someone can read in 10 seconds.)
✍️ 1📣 1
📄
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
Jonas Geiping, Sean McLeish, Neel Jain, John Kirchenbauer,... · 2025
(To be written by the class after week 3.)
no posts yet, be the first

Latest posts

📣 Communicate to the world Oct 21, 2026
Matryoshka Language Model Suites in a thread
Social media posts for Matryoshka Language Model Suites.
Matryoshka LM suites
✍️ Writing Oct 21, 2026
Editing the methods of </div>
sub-models are made of blocks
Matryoshka LM suites
</a>
🔬 Research skills Oct 20, 2026
critique of matryoshka paper (GA26)
The narative about saving on parameters is diluted by a secondary point about cheap distiliation.
📣 Communicate to the world Oct 16, 2026
Week 0 in one page: tokens keep knowledge apart
The paper, what we learned, and everything the class wrote about it, in a five-minute read.
Cross-lingual knowledge in pretraining
✍️ Writing Oct 16, 2026
Three things to remember, and where the paper hides them
The points most readers should leave with, and how the paper could make them harder to miss.
Cross-lingual knowledge in pretraining
📣 Communicate to the world Oct 16, 2026
The paper in four feeds
Posts for X, Bluesky, Mastodon and LinkedIn, and why each one is written differently.
Cross-lingual knowledge in pretraining
 </div>

All posts →

About the course

Every week we read one paper. Before class, each participant picks a task: rewrite the abstract, critique the paper, redo a graph, reimplement an experiment, write the social media thread, find the papers it builds on, and more. Then we write it up here, so the work is useful to everyone who reads the paper after us.

You are welcome to experiment. Do the default thing, or try a new way that might fail. We encourage failure: learning only happens when you are exposed to the unknown.

Want to join? See the guide. Reading along from outside the class? Comments, corrections and follow-ups are welcome on GitHub Discussions.

</div>