QUIVER defines sensitivity matrices with type-dispatched metrics, trajectory divergence decomposition, bifurcation thresholds, and distribution faithfulness to analyze perturbation effects in LLM pipelines, validated on enterprise and public systems with over 8200 traces.
Heuristics include: gap analysis that returned marginal results, confidence scores near decision thresholds, or categorical classifications with low model confidence
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
QUIVER: A Formal Framework for Quantifying Perturbation Propagation and Bifurcation in Compound AI Systems
QUIVER defines sensitivity matrices with type-dispatched metrics, trajectory divergence decomposition, bifurcation thresholds, and distribution faithfulness to analyze perturbation effects in LLM pipelines, validated on enterprise and public systems with over 8200 traces.