RegressionDetective
An interactive tutor that walks a stats or ML student through a linear regression problem two ways at once: the classical closed-form OLS derivation on one side, and a live visualization of a transformer learning the same fit in-context on the other.
Undergraduate statistics and machine learning students
- Side-by-side derivation: closed-form equations next to transformer attention maps
- Step-through mode that pauses the transformer's iteration and explains what each layer is doing
- Auto-generated practice problems where students pick which method a given dataset favors
- Exportable notebook that links the student's scratch work to the underlying paper
Papers like Hagiwara's are making it concrete that transformers actually implement closed-form math, giving educators a teachable bridge between classical stats and modern AI that students have been missing.
Free OLS interactive tools (PhET, U Toronto, linear-regression.dev) get classroom use, and Georgia Tech's Transformer Explainer reached 500K+ learners, confirming an audience for stats/ML interactive tutors, but no direct search signal for a tool specifically bridging OLS with transformer ICL.Transformer Explainer Shows How AI is More Math Than Human ↗Least-Squares Regression - Linear Regression - PhET Interactive Simulations ↗
Side-by-side comparison of classical OLS derivation AND a live transformer fitting the same regression is not found; existing tools split into OLS visualizers (PhET, U Toronto, ealebrahim) or transformer explainers (Polo Club's, LLM Visualization, TransforLearn) with no bridge.Transformer Explainer: LLM Transformer Model Visually Explained ↗Ordinary Least Squares Interactive - ealebrahim.com ↗
Khanmigo is $4/mo individual with paid school/district licensing and Brilliant Premium exists for STEM, showing willingness to pay for interactive tutoring, but consumer-grade stats tools mostly free, capping price ceiling.Khan Academy Cost Guide And Pricing 2026 - softwarefinder.com ↗Pricing & Plans - Help Center | Brilliant ↗
Hagiwara's arXiv paper (2025) and the broader 'transformers implement closed-form math' line of work provides a durable pedagogical insight; risk that the specific linear-self-attention framing will need updates as transformer architectures evolve.In-context learning of closed form solution to simple linear regression using transformer with linear self-attention ↗
Building a small in-context-learning transformer on linear regression is a well-replicated experiment and pairing it with off-the-shelf OLS derivation plotting is straightforward; prior art (PhET, Transformer Explainer) shows the interactive viz stack is solvable.Transformer Explainer: Interactive Learning of Text-Generative Models ↗Linear Regression Visualizer ↗