DiscoveryRSI: research hypotheses with an inspectable investigation plan

Development · 2026-10-16

DiscoveryRSI releases inspectable scientific hypothesis trees

Public code and model artifacts connect proposed claims with investigations, decision criteria and recorded lineage.

Why it matters

This gives research teams a concrete structure to evaluate and revise AI-generated experimental plans.

What to watch

Watch for independent reproductions, documented failed hypotheses and tests outside the release’s life-science domains.

Still uncertain. The release’s expert evaluation is preliminary. Scientific validity still depends on real experiments and domain review.

Impact assessment reviewed 2026-10-19.

How it works

Each hypothesis can carry a feasible investigation, decision criteria and alternative explanations. The release records lineage and uses a fixed evaluation harness. Its life-science evaluations combine automated scoring with preliminary expert review.

DiscoveryRSI implementation and evaluation ↗DiscoveryRSI paper ↗DiscoveryRSI paper listing ↗

Trace one claim to an experiment

Choose one public hypothesis tree and trace its central claim to an observable result that could reject it. Record missing data, equipment and controls before treating the proposal as an executable experiment.

DiscoveryRSI implementation and evaluation ↗DiscoveryRSI paper ↗DiscoveryRSI paper listing ↗

What you can do

1. Read the linked original and compare its documented scope with the task you need to complete. Open reference ↗

Where the evidence stops

A well-formed proposal has not thereby been experimentally validated. No model, demo or laboratory experiment was run during this review.

Sources read. Sources read October 19, 2026. No installation, inference or complete external workflow was executed.

Sources & further reading