Pre Registration Python Packages

fieldtrial

Statistically rigorous real-world evaluation for robot policies. Plan how many rollouts you need, run them randomized and blinded from a phone at the robot, let fieldtrial switch LeRobot or openpi policies itself, and get a report that only says "better" when a pre-registered test does.

claimbound-evidence

Evidence cards for narrow public AI, ML and data claims: frozen protocol, hashed sources, honest result status.

thesis-ci

CI for investment theses: write the tests that would prove a thesis wrong in YAML, lint a research archive so every number cites a filing and public files carry no valuation or advice, freeze each quarter's thresholds once results are out, and Brier-score forecasts. Python CLI and GitHub Action. 中文版: zh-CN/

falsify

Fix the pass bar before the test runs. PRML is an open spec and CLI that hashes and timestamps evaluation criteria (SHA-256, RFC 3161) so anyone can check them later. Python, JS, Go, Rust.

falsify-inspect

PRML pre-registration adapter for Inspect AI eval logs. Listed on the UK AI Safety Institute's Inspect extensions page. MIT.

nullbar

Pre-registration and honest statistics for trading research: trial ledgers, deflated Sharpe, clustered inference, leak detection, fill realism.

mlflow-falsify

Fix the pass bar before the test runs. PRML is an open spec and CLI that hashes and timestamps evaluation criteria (SHA-256, RFC 3161) so anyone can check them later. Python, JS, Go, Rust.

falsify-giskard

PRML pre-registration for Giskard scenario results: commit an eval claim to a SHA-256 before the run, then verify the result against it.

pen-compare

Hierarchical certification framework (5-gate TrueWriterScore) for programmable genome editors, with sensitivity analysis, cross-pipeline triangulation, and a RAG Q&A layer over the PEN-STACK.