Saved in:
Bibliographic Details
Main Authors: Wu, Yunnan, Chen, Paul, Baranwal, Deshank, Zhou, Jinlong, Yuan, Jian
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.21036
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909554400296960
author Wu, Yunnan
Chen, Paul
Baranwal, Deshank
Zhou, Jinlong
Yuan, Jian
author_facet Wu, Yunnan
Chen, Paul
Baranwal, Deshank
Zhou, Jinlong
Yuan, Jian
contents We present an agentic framework, Thinker, which achieves state of art performance in challenging reasoning tasks for realistic customer service scenarios that involve complex business logic and human interactions via long horizons. On the $τ$-bench retail dataset, Thinker achieves 82.6\% success rate with GPT-4o (version 2024-06-01) (baseline: 68.3\%), and 81.9\% success rate with Llama-3.1 405B (baseline: 49.6\%), without any fine-tuning. Thinker effectively closes the gap in reasoning capabilities between the base models by introducing proper structure. The key features of the Thinker framework are: (1) State-Machine Augmented Generation (SMAG), which represents business logic as state machines and the LLM uses state machines as tools. (2) Delegation of tasks from the main reasoning loop to LLM-powered tools. (3) Adaptive context management. Our prompting-only solution achieves signficant gains, while still maintaining a standard agentic architecture with a ReAct style reasoning loop. The key is to innovate on the tool interface design, as exemplified by SMAG and the LLM-powered tools.
format Preprint
id arxiv_https___arxiv_org_abs_2503_21036
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Art of Tool Interface Design
Wu, Yunnan
Chen, Paul
Baranwal, Deshank
Zhou, Jinlong
Yuan, Jian
Artificial Intelligence
We present an agentic framework, Thinker, which achieves state of art performance in challenging reasoning tasks for realistic customer service scenarios that involve complex business logic and human interactions via long horizons. On the $τ$-bench retail dataset, Thinker achieves 82.6\% success rate with GPT-4o (version 2024-06-01) (baseline: 68.3\%), and 81.9\% success rate with Llama-3.1 405B (baseline: 49.6\%), without any fine-tuning. Thinker effectively closes the gap in reasoning capabilities between the base models by introducing proper structure. The key features of the Thinker framework are: (1) State-Machine Augmented Generation (SMAG), which represents business logic as state machines and the LLM uses state machines as tools. (2) Delegation of tasks from the main reasoning loop to LLM-powered tools. (3) Adaptive context management. Our prompting-only solution achieves signficant gains, while still maintaining a standard agentic architecture with a ReAct style reasoning loop. The key is to innovate on the tool interface design, as exemplified by SMAG and the LLM-powered tools.
title The Art of Tool Interface Design
topic Artificial Intelligence
url https://arxiv.org/abs/2503.21036