research-document

Domain Leakage Report

Domain Leakage Report

Packet Provider Introduced Terms Classification Notes
ECR-000003-EXP001-P001A gpt none generic_procedural_language low
ECR-000003-EXP001-P001A claude evidence possible_structural_inference Low. No domain terminology was introduced. Structural observations were derived solely from graph edges and node notes. The recognized pattern is broad enough that it maps to multiple procedure families; no specific domain structure was imported.
ECR-000003-EXP001-P001A gemini none generic_procedural_language none
ECR-000003-EXP001-P001B gpt none generic_procedural_language low
ECR-000003-EXP001-P001B claude none generic_procedural_language low; no domain terms were introduced; structural observations are grounded solely in graph edges and node notes; the parallel-path-with-reeval topology is generic enough that no specific artifact was assumed
ECR-000003-EXP001-P001B gemini none generic_procedural_language generic_procedural_language
ECR-000003-EXP001-P001C gpt none generic_procedural_language low
ECR-000003-EXP001-P001C claude diagnosis; evidence likely_prior_knowledge_leakage Low-to-moderate. The resemblance noted in recognized_artifact could bias interpretation toward diagnosis-style semantics; extraction was restricted to packet vocabulary (paths, differentiating input, re-sorting, separation) and no domain terminology was imported into any layer.
ECR-000003-EXP001-P001C gemini none generic_procedural_language none
ECR-000003-EXP001-P001D gpt none generic_procedural_language low
ECR-000003-EXP001-P001D claude diagnosis; hypothesis testing likely_prior_knowledge_leakage low-to-moderate: the familiar 'generate-then-discriminate' shape creates a temptation to import domain framing (e.g., hypothesis testing or diagnosis vocabulary); this was avoided by keeping all labels at the packet's own level of abstraction (paths, differentiating input, fit, separation).
ECR-000003-EXP001-P001D gemini none generic_procedural_language generic_procedural_language

Leakage classifications are observational and may still reflect structural inference rather than memorized framework recall.