LIVE RUN
No runPipeline progress
Launch a pipeline to see stage-level progress.
RESEARCH OPERATIONS
Evidence pipeline
Connecting
CHILDSAFE EVIDENCE PIPELINE V2
Literature to framework,
Literature to framework,
fully traceable.
Run four-engine discovery, independent Agent screening, Dynamic Drop, exact-span verification and candidate rubric synthesis from one control surface.
No active pipeline
EVENT STREAM
Latest activity
No pipeline events yet.
RUN HISTORY
Jobs
| Run | Status | Created | Attempt | Mode | Families | Actions |
|---|
METHOD AT A GLANCE
Agentic evidence-to-framework workflow
The seed taxonomy scopes discovery; Agents make bounded semantic decisions, deterministic systems enforce reproducibility, and every published field retains a path to exact source evidence.
LLM Agent
Deterministic system
Verification gate
Research output
01Scope & discoverDefine what to search
INPUTSeed taxonomy10 risk families provide scope, synonyms and starting boundaries.
AGENTQuery PlannerGenerates versioned, engine-specific youth and protection queries.
SYSTEMLiterature retrievalOpenAlex, Semantic Scholar, Crossref and arXiv Top 100; Scholar checks coverage.
SYSTEMFilter, deduplicate & rank2023+ filter, DOI/title matching, RRF and citation-aware priority.
02Screen & stopDecide what deserves full text
2 AGENTSIndependent screeningPopulation-first checks youth relevance; Risk-first checks mechanism fit.
AGENTAdjudicatorRuns only for disagreement, uncertainty or a probability gap above 0.25.
GATEDynamic DropDetects a dramatic relevance decline, then challenges it with engine-tail and Scholar sentinels.
SYSTEMFull-text queueOnly eligible open-access or locally authorized documents advance.
03Verify & buildTurn evidence into auditable rules
AGENTEvidence LocatorFinds exact page, section, offsets and short verbatim spans.
GATEClaim VerifierChecks quote reproduction, entailment, age, sample and causal language.
AGENTSynthesis AgentMaps verified claims, records conflicts and proposes taxonomy updates.
AGENTRubric BuilderCreates age-specific anchors, safe-response rules and hard-fail candidates.
GATERelease gatesRequires complete lineage, reproducible quotes and no unsupported propagation.
OUTPUTCandidate releaseTaxonomy, rubric cards, ontology, evidence ledger and trace report.
END-TO-END LINEAGERule → Statement → Claim → Span → Page → Source → Query
FAILS A GATEQuarantine with reason; never silently propagated
VERSION CONTROLFormal taxonomy and rubric v1 remain frozen
DISCOVERY
Retrieval coverage
Top-100 coverage is shown before eligibility screening. Missing engines remain visible.
Select a run to inspect engine coverage.
Source funnel
No retrieval data.
Dynamic Drop
No cutoff decisions.
MULTI-AGENT REVIEW
Screening decisions
Population-first and Risk-first decisions are independent; material disagreement invokes adjudication.
Ranked papers and Agent probability
| Rank | Family | Paper | Origin | RRF | Citation | Agent p | Decision |
|---|
PROVENANCE
Evidence explorer
Public evidence is limited to metadata and verified excerpts of at most 280 characters.
Trace an entity
Enter an entity ID or open a paper from the screening table.
OUTPUTS
Candidate release
Formal taxonomy and rubric v1 remain unchanged. Public downloads are sanitized allowlisted artifacts.
No completed release selected.
Artifacts
No artifacts available.