KX-08 · Validation
Helios Multi-Agent Scientific Council
Run a structured multi-perspective review that surfaces disagreement rather than averaging it.
HypothesisModel-Inferred
A council of role-conditioned analytical passes — experimentalist, theorist, metrologist, statistician, adversary, and compliance reviewer — each producing an independent position before any synthesis.
Institute module developed within the KRYOS v6 framework ecosystem. Established scientific methods retain their conventional attribution.
When to use
Before an evidence-state elevation, funding gate, or external communication.
When not to use
As a replacement for human peer review or institutional sign-off.
Required inputs
- Evidence register
- Artifact elimination report
- Claim under review
Optional inputs
- Prior review findings
- Competing interpretations
Assumptions and constraints
- Disagreement is preserved and reported, not resolved by majority.
Process steps
- 01Run each role pass independently against the same evidence.
- 02Record each position with its supporting claim IDs.
- 03Identify points of disagreement and their evidentiary basis.
- 04Produce a synthesis that preserves minority positions.
- 05Route unresolved disagreement to human reviewers.
Human review gates
- Named human reviewers own the final position; the council output is advisory.
Primary output — Council review with preserved dissent and routed open questions.
- role, position, supporting_claims[], confidence_statement
- disagreement_points[] with evidentiary basis
- routed_questions[] with human owner
Failure signals
- All roles agree without any identified open question.
Quarantine triggers
- Council output presented as peer review.
Mini prompt
Review [CLAIM] from six independent positions: experimentalist, theorist, metrologist, statistician, adversary, compliance. Report each position separately with supporting claim IDs. Preserve disagreement and route open questions to a named human owner.