Documentation

Think-Reason-Learn (TRL) is the decision harness for predictions. Install it, fit a model on a few examples, and read the reasons behind every call.

pip install "git+https://github.com/vela-research/think-reason-learn.git"Copy

Start here

1InstallPython 3.13 or newer, from GitHub. 2Quick startFit a small GPTree on six founder profiles. 3API referenceEvery class, setting and method.

Methods

Question treeGPTreeDecision trees whose questions a language model writes. Question ensembleRandom Rule ForestMany yes-or-no questions combined into one transparent model. Written policiesPolicy InductionShort policies learned from examples, plain enough to check. Weighted rulesReasoned Rule MiningIf-then rules with calibrated confidence, cheap checks first. Next best questionVerifiable RLLearns what to check next, and stops when it is sure. Rules as codeFeature GenerationRules written as small Python functions that run without a model.

ChangelogContributingLicencePapersGitHub