The archive
Research results and the fundamentals behind the models.
MIT associate professor Christina Delimitrou argues that cloud infrastructure wastes enormous capacity, and that machine learning can reclaim it. The profile is light on hard numbers.
5 min read
An MIT profile of associate professor Cathy Wu traces years of failed RL experiments to a 2023 result: train on the 10% of problems that generalize, and skip the rest. Up to 30x more efficient.
5 min read
Anthropic's new robot exposure index finds today's machines can perform 74% of physical US job tasks, but are cost-competitive on just 0.3% of work. Economics beats capability.
5 min read
AWS walks through GRPO post-training of Qwen3-VL-8B on SageMaker HyperPod with SkyRL, reporting a maze solve rate climb from 43.75% to 96.875% on a 64-maze evaluation set.
6 min read
Researchers at MIT's McGovern Institute used an LLM to build a 49-factor risk lexicon, then trained a small, interpretable model on roughly 16,000 crisis conversations.
5 min read
A workshop recap from Ai2 and Providence Swedish names five obstacles to AI-driven discovery, from scientific taste to the risk of amplifying flawed methods.
5 min read
MIT and Harvard researchers report xvr, a method that aligns live surgical X-rays with a patient's 3D scan in seconds, trained per patient in about five minutes.
5 min read
MIT researchers report a deployment-time method that forces pretrained flow-matching models to satisfy hard safety constraints, with no retraining required.
5 min read
What embeddings are, how vector spaces capture similarity, and how teams use them for search, recommendations, clustering, and deduplication.
8 min read
Why a model can look excellent in training and fail on real data, how to spot overfitting early, and the practical techniques that keep models honest.
8 min read
The data problems that quietly sink machine learning projects, from label noise and leakage to drift, and a practical checklist for catching them early.
8 min read
A plain-language tour of the transformer: how attention works, why position has to be added in, and what each layer contributes to the final result.
8 min read