Design Behaviour Codes (DBCs): A Taxonomy-Driven Layered Governance Benchmark for Large Language Models

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• The DBC layer reduces aggregate Risk Exposure Rate (RER) from 7.19% to 4.55%, representing a 36.8% relative risk reduction.result 95%
• The DBC framework achieves substantially higher risk reduction (36.8%) compared to a standard safety moderation prompt (0.6%).result 90%
• The DBC system enables EU AI Act compliance scoring of 8.5/10 through automated evaluation.result 90%
• Graybox adversarial attacks achieve a DBC Bypass Rate of 4.83%.result 90%
• The Integrity Protection cluster (MDBC 081-099) delivers the highest per domain risk reduction.result 85%
• The DBC framework uses a three-arm controlled design (Base, Base plus Moderation, Base plus DBC) for causal attribution of risk reduction.method 85%
• The evaluation protocol achieves substantial inter-rater agreement with Fleiss kappa > 0.70.method 90%
• The DBC system is model-agnostic, jurisdiction-mappable, and auditable unlike training-time alignment methods.technical 80%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card