TRACER: Trace-Based Adaptive Cost-Efficient Routing for LLM Classification

TL;DR

TRACER is an open-source system that uses production logs to train surrogates for LLM classification, optimizing when to deploy them based on agreement thresholds to reduce inference costs.

Contribution

It introduces a trace-based adaptive routing system that determines surrogate deployment boundaries and provides interpretability artifacts for LLM classification tasks.

Findings

Achieves 83-100% surrogate coverage on intent benchmarks.

Fully replaces the teacher on a 150-class benchmark.

Correctly refuses deployment when representations are unreliable.

Abstract

Every call to an LLM classification endpoint produces a labeled input-output pair already retained in production logs. These pairs constitute a free, growing training set: a lightweight surrogate trained on them can absorb a significant portion of future traffic at near-zero marginal inference cost. The open questions are when the surrogate is reliable enough to deploy, what it handles versus defers, and how that boundary evolves as data accumulates. We introduce TRACER (Trace-based Adaptive Cost-Efficient Routing), an open-source system that trains ML surrogates on an LLM's own production traces and governs deployment through a parity gate: the surrogate is activated only when its agreement with the LLM exceeds a user-specified threshold {\alpha}. To make the routing boundary transparent, TRACER generates interpretability artifacts describing which input regions the surrogate handles, where it plateaus, and why it defers. On a 77-class intent benchmark with a Sonnet 4.6 teacher, TRACER achieves 83-100% surrogate coverage depending on the quality target {\alpha}; on a 150-class benchmark, the surrogate fully replaces the teacher. On a natural language inference task, the parity gate correctly refuses deployment because the embedding representation cannot support reliable separation. The system is available as open-source software.