Reinforcing Real-world Service Agents: Balancing Utility and Cost in Task-Oriented Dialogue

Ning Gao, Wei Zhang, Yuqin Dai, Ling Shi, Ziyin Wang, Yujie Wang, Wei He, Jinpeng Wang, Chaozheng Wang 🖂 · 2026

Read Reinforcing Real-world Service Agents: Balancing Utility and Cost in Task-Oriented Dialogue by Ning Gao, Wei Zhang, Yuqin Dai, Ling Shi, Ziyin Wang, Yujie Wang, Wei He, Jinpeng Wang, Chaozheng Wang 🖂 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-22697.txt.

Description

Ning Gao and eight co-authors balance kindness and cost in real-world service agents. Language-model agents now handle full tasks, but marrying empathetic talk with budget-aware choices is still open. Existing methods miss the strategic trade-off between pleasing users and controlling spend.

InteractCS-RL reframes task dialogue as multi-granularity reinforcement learning. A user-centred interaction framework provides a high-fidelity training gym where agents explore strategies against persona-driven users. Cost-aware multi-turn policy optimization then steers learning with hybrid advantage estimates, generative process credits, and a PID-Lagrangian cost controller toward the Pareto edge between reward and cost. Tests on real business scenarios and tool-agent-user benchmarks beat baselines on all three evaluation fronts.

Catalog record

CREATORNing Gao, Wei Zhang, Yuqin Dai, Ling Shi, Ziyin Wang, Yujie Wang, Wei He, Jinpeng Wang, Chaozheng Wang 🖂
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-22697

Browse the full library · About the Tower of Babel