RSI4Science
English

This page is available in English. Choose a language to open its dataset guide.

Benchmarks and demos

Data-building strategies · proposed setting

Build more informative molecular test cases

Explore how a system could improve the way it selects and constructs property-evidence cases, then use that strategy to assess later changes to itself. A source-backed MolP2SBench teaching pair illustrates property feedback; a recursive case builder is still proposed.

Start with a real diagnostic example

Try a MolP2SBench source–counterfactual pair and inspect its feedback. The linked pilot proposes a revisable molecular solver and improver; it does not implement the case-building research direction below.

Open the molecular demo

Your first contribution

You can begin with a source, dataset, method or diagnostic. Researchers supply and run any proposed RSI system in their own workflow.

Scenario proposal and review criteria

For a complete research scenario

A future study would also need a recoverable workspace, scientific feedback and independent improvement tests. A source listing can be reviewed separately from that study.

Study protocol · Available data

Proposed research question

Can a revised case-building strategy help a system make better subsequent self-modifications, while preserving independent chemical validity and evaluation?

Where recursion would happen

Evidence selection, case construction and coverage diagnostics are candidate editable capabilities. The installed strategy must be used to test or design later system changes, including further changes to that strategy.

Generating more examples or updating a fixed learner is insufficient. This proposal still needs a credible scientific limitation, an executable case builder and evidence that its revisions participate in later self-modification.

Scientific feedback and evaluation

A source-backed candidate pool would supply structures, property evidence, units and provenance. Measured properties, computed proxies and constructed cases must remain distinct. Missing observations cannot be treated as negative results.

Keep chemical validity checks and final scientific outcomes outside the editable builder. Group related scaffolds, sources and parent records before splitting. Test whether a new improver produces better scientific methods in fresh workspaces, rather than merely easier cases or higher self-assigned scores.

The comparison we need

Use fixed construction, history-only selection and budget-matched search controls with the same accessible source evidence. If an exact filter or strong fixed method solves the proposed setting, record that limit rather than assuming RSI is needed.

Read the shared evaluation design.

Available now

  • One attributed MolP2SBench source–counterfactual teaching pair
  • A diagnostic demo and proposed feedback-to-RSI protocol; no implemented case builder

Needed for a future study

  • A source with usable evidence, rights and scientifically meaningful grouping
  • A reasonable starting case builder and observed performance records
  • Independent validity checks, future improvement tests and controlled runs

A bounded contribution · Data curation and scientific test design

Propose a molecular case-building problem

This scope is for a research scenario proposal. To contribute a dataset or source alone, use the dataset contribution form; a complete RSI workspace is optional.

What to deliver

  1. Provide a source and version, reuse terms and a representative candidate or condition with property provenance.
  2. Identify a recoverable selection or case-building method and any documented coverage limitation. Separate observations from hypotheses.
  3. Suggest an independent validity check and grouping boundary; explain how a revised construction strategy could participate in later self-modification.

How it will be reviewed

A domain reviewer checks that the evidence supports the scientific question and that constructed cases can be validated independently. A resource-only proposal is welcome; it is not yet a complete RSI scenario.

Sources and contributors

The listing distinguishes original source authors from contributors who curate data, build software or design new checks. Preserve the source citation and record the contributed version.

Source and reuse notes.

Propose a scientific task

Agree a manageable scope and a reviewer before starting. The form saves a draft on your device; contact the project lead to discuss it.

Project-led design. No independent scientific approval or completed RSI experiment is claimed. Accepted contributions and review decisions will be recorded as that work occurs.