llm-belief-pipeline-is-fully-quality-enforced
OUT derived (depth 4)
The system's complete LLM-driven belief pipeline — both generation (derive with safety, completeness, and coverage) and classification (list_negative with defensive bounding) — achieves fully code-enforced quality constraints at every stage, provided minimum-antecedent validation moves from prompt-only to code-enforced.
Summary
Both the belief generation and negative-classification pipelines have strong quality safeguards, but one gap remains: the check that derived beliefs have enough supporting antecedents is currently enforced only by the LLM prompt, not by code. Until that validation is hardened into actual code enforcement, the pipeline cannot be considered fully quality-controlled at every stage.
Justifications
SL — Full LLM belief pipeline quality requires code-enforced antecedent validation
Antecedents (all must be IN):
- derive-pipeline-has-complete-coverage — The derive pipeline achieves complete coverage along three axes: safety (fail-soft validation, Jaccard retraction guards, environment isolation), completeness (exhaustive exploration with guaranteed termination), and production-readiness (accurate proportional budgets, roundtrippable prompt format).
- list-negative-is-defensively-bounded — The negative belief listing pipeline applies defense-in-depth: keyword pre-filtering narrows candidates before LLM classification, hallucinated node IDs are discarded against the actual network, and malformed LLM output falls back gracefully to zero count rather than raising.
Unless (any of these IN defeats this justification):
- derive-min-antecedents-is-prompt-only — The minimum-2-antecedents rule for derived beliefs is enforced only by the LLM prompt instructions, not validated in code by `validate_proposals`.