What question did this study set out to answer?

The aim is to explore a measurement framework for multi-model orchestration while ensuring auditable decision-making.

February 2, 2026Open Access

ACAR: Adaptive Complexity Routing for Multi-Model Ensembles with Auditable Decision Traces

Key Points

The aim is to explore a measurement framework for multi-model orchestration while ensuring auditable decision-making.
Developed ACAR as a measurement framework for multi-model ensembles.
Used self-consistency variance from probe samples for task routing.
Evaluated across 1,510 tasks and produced over 7,550 auditable runs.
σ-based routing achieved 55.6% accuracy, surpassing the two-model baseline of 54.4%.
Avoided full ensembling on 54.2% of tasks, demonstrating efficiency.
Retrieval augmentation led to a 3.4% accuracy drop due to noise from misaligned experience.

Abstract

We present ACAR (Adaptive Complexity the mechanism is modelagnostic and requires no learned components. What does not hold: (1) Retrievalaugmentation decreased accuracy by 3.4 percentage points—median retrieval similarity was only 0.167, demonstrating that experience injection without semanticalignment introduces harmful noise rather than grounding. (2) When models agreeon incorrect answers (σ=0), no downstream ensemble can recover; this “agreementbut-wrong” failure mode is intrinsic to self-consistency and bounds achievableaccuracy at 8pp below full ensembling. (3) Attribution estimates based on proxysignals (response similarity, entropy) showed weak correlation with ground-truthleave-one-out values; practical attribution requires explicit counterfactual computation. This paper documents what assumptions fail in practice, providing falsifiablebaselines for future work on routing, retrieval, and multi-model attribution.

Ask AI

Helpful

Bookmark

View Full Paper