Interactive research methods · Verified + experimental
Don’t just read
the method. Run it.
Explore three verified algorithm laboratories, or experimentally match an arXiv paper’s central method to the exact trusted experience that can explain it faithfully.
01 / The learning gap
A summary tells you what. A laboratory shows you why.
Research methods are written to be precise, not necessarily tangible. Static explanations flatten an iterative process into another block of text. The crucial intuition often lives in the transition between steps.
Here, every control maps to actual algorithm logic. Advance a single phase, inspect the state, then form your own mental model.
Start from a verified foundational paper.
Manipulate a real input to the algorithm.
See state, motion, and objective values evolve.
02 / Verified collection
Three methods. One careful lab at a time.
Move from geometric clustering to informed search and the mechanics of attention. Every laboratory runs verified, repository-owned algorithm code.
k-Means Clustering
James MacQueen, 1967
Step through assignment and centroid updates while tracking how cluster geometry changes inertia.
A* Search
Peter E. Hart, Nils J. Nilsson, and Bertram Raphael, 1968
Observe how the heuristic guides frontier expansion toward a minimum-cost path.
Scaled Dot-Product Attention
Ashish Vaswani et al., 2017
Connect score scaling and normalization to the attention pattern and resulting value mixture.
A deliberate boundary
Each laboratory couples deterministic algorithm code with a trusted, purpose-built visualization. Experimental analysis returns schema-validated data that may still be imperfect; no model-generated code is compiled or executed.