arXiv:2605.17482v2 Announce Type: replace-cross Abstract: Local XAI audits compare a finite block of learned vectors with a weak side signal. Baselines such as nearest-neighbor lookup, low-rank coordinate models, and relation factorization expose different parts of this audit. We introduce Relational Semantic Decomposition, abbreviated as RSD, as a local triangulation audit for learned vector blocks. Given coordinates X and a declared bounded weak affinity proxy A, RSD fits simplex memberships S and coordinate poles C. It reuses S in a relation decoder for A and reports the coordinate residual
The proliferation of complex learned vector blocks in AI systems necessitates advanced auditing techniques to ensure transparency and reliability, pushing for innovations in explainable AI (XAI).
This development offers a novel method for auditing learned vector blocks, enhancing the explainability and interpretability of AI, which is crucial for deployment in sensitive applications and for regulatory compliance.
The introduction of RSD provides a new, more robust primitive for local XAI audits, moving beyond simpler baselines and enabling deeper inspection of AI model decision-making.
- · AI developers
- · Auditors and regulators
- · Companies deploying AI
- · Explainable AI researchers
- · Black-box AI systems
- · Organizations with opaque AI models
Improved understanding and trust in complex AI systems, especially those using vector embeddings.
Faster adoption of AI in risk-sensitive sectors due to enhanced auditability and explainability.
Potential for new industry standards in AI auditing and compliance, favoring methods like RSD over current baselines.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG