Profile-Then-Reason: Bounded Semantic Complexity for Tool-Augmented Language Agents

TL;DR

This paper introduces Profile-Then-Reason (PTR), a structured framework for tool-augmented reasoning in language models that reduces latency and error propagation by explicit workflow synthesis and bounded verification.

Contribution

The work presents a novel bounded execution framework for structured reasoning, limiting language model calls and improving accuracy on certain benchmarks.

Findings

PTR achieves exact-match in 16 of 24 configurations against ReAct baseline.

Particularly effective on retrieval-centered and decomposition-heavy tasks.

Reactive execution remains preferable for online adaptation tasks.

Abstract

Large language model agents that use external tools are often implemented through reactive execution, in which reasoning is repeatedly recomputed after each observation, increasing latency and sensitivity to error propagation. This work introduces Profile--Then--Reason (PTR), a bounded execution framework for structured tool-augmented reasoning, in which a language model first synthesizes an explicit workflow, deterministic or guarded operators execute that workflow, a verifier evaluates the resulting trace, and repair is invoked only when the original workflow is no longer reliable. A mathematical formulation is developed in which the full pipeline is expressed as a composition of profile, routing, execution, verification, repair, and reasoning operators; under bounded repair, the number of language-model calls is restricted to two in the nominal case and three in the worst case. Experiments against a ReAct baseline on six benchmarks and four language models show that PTR achieves the pairwise exact-match advantage in 16 of 24 configurations. The results indicate that PTR is particularly effective on retrieval-centered and decomposition-heavy tasks, whereas reactive execution remains preferable when success depends on substantial online adaptation.