🤖 AI Summary
Existing Shapley value–based attribution methods are impractical in composite AI systems due to their requirement to evaluate all subsets of components—a prohibitive task when systems involve unobservable third-party APIs or centralized routing. To address this, this work proposes BOHM, a novel approach that leverages inherent system routing weights to construct a hierarchical attribution tree, enabling multi-granular contribution analysis through the product of root-to-leaf path weights. BOHM achieves zero marginal computational cost and requires no internal system access, offering, for the first time, zero-cost, multi-resolution attribution without re-evaluating the system. Evaluated across 18 language models, 5 drivers, and U.S. Census data, BOHM attains a Kendall’s τ of 0.928—comparable to SHAP’s 0.980 but at 9,000× lower computational overhead—and recovers ground-truth rankings with τ up to 0.722 across resolution levels.
📝 Abstract
Compound AI systems route tasks through hierarchies of specialised components. Attribution is dominated by Shapley-based methods (SHAP), which decompose a coalition value function into per-component marginal contributions and require evaluation of the system on arbitrary component subsets. That requirement fails for third-party APIs, opaque endpoints, and agentic orchestrators that concentrate routing on a few tools, leaving most coalitions un-evaluable from the deployed orchestrator. We introduce BOHM, which extracts a hierarchical attribution tree directly from the routing weights such systems already maintain: leaf attribution is the path product of root-to-leaf routing weights; level-k attribution is the induced distribution over depth-k nodes. The method has zero marginal cost, requires no access to component internals, and provides multi-resolution attribution at every level simultaneously, which flat methods cannot offer at any evaluation budget. BOHM and SHAP answer different questions and converge when the deployed router routes near-optimally. On 18 LLMs in a 3-level hierarchy over 880 LiveCodeBench problems, BOHM yields Kendall tau=0.928; SHAP reaches tau=0.980 at 9,000x more coalition evaluations per seed. On a 5-driver, 7-benchmark agentic study (35 cells, complete coverage), drivers concentrate routing on a single tool (top-share median 0.65), and cell-level tau(BOHM,SHAP) is predicted by whether the driver's top pick is the empirically best tool (mean +0.22 vs ~+0.01). On a US Census hierarchy (475 leaves, 4 levels), BOHM recovers ground-truth rankings at every level (tau up to 0.722). BOHM satisfies efficiency, monotonicity, symmetry, and weak suppression but not Shapley's additivity. It is best understood as a complementary primitive: a multi-resolution decomposition computable wherever routing state exists, whose disagreement with Shapley is itself diagnostic.