Skip to content
Chimeraforge

Reinforcement Learning Projects

Environments where a program chooses one move at a time and is scored on how the task ends, and the tests that compare one strategy with another on identical cases.

  • 4 projects in this collection
  • Runs in your browser

Projects by collection

Flight Routing Lab

With a tight deadline, planning ahead gets 40 of 64 simulated trips to JFK on time where booking the earliest nonstop gets 0, but 8 fewer arrive at all.

Sample data

Offline Policy Evaluation Lab

A new decision policy looks 0.27 better on records of past decisions, but a control that ignores the situation gets 65% of that, and a changed reward erases it.

Sample data

Customer Service Environment

Two scripted runs refund the same $48.00 on a double-charged order: refunding the duplicate charge scores 1.00, refunding the real one scores −0.40.

Sample data

Code Verification Lab

Four candidate bug fixes: a two-test smoke suite passes three, the full six-test suite passes one. Passing the bug’s own example is not a fix.

Sample data