Preprint · 2025

$τ^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment

Victor Barres, Honghua Dong, Soham Ray, Xujie Si, Karthik Narasimhan

Abstract

Existing benchmarks for conversational AI agents simulate single-control environments, where only the AI agent can use tools to interact with the world, while the user remains a passive information provider. This differs from real-world scenarios like technical support, where users need to actively participate in modifying the state of the (shared) world. In order to address this gap, we introduce $τ^2$-bench, with four key contributions: 1) A novel Telecom dual-control domain modeled as a Dec-POMDP, where both agent and user make use of tools to act in a shared, dynamic environment that tests both agent coordination and communication, 2) A compositional task generator that programmatically creates diverse, verifiable tasks from atomic components, ensuring domain coverage and controlled complexity, 3) A reliable user simulator tightly coupled with the environment, whose behavior is constrained by tools and observable states, improving simulation fidelity, 4) Fine-grained analysis of agent performance through multiple ablations including separating errors arising from reasoning vs communication/coordination. In particular, our experiments show significant performance drops when agents shift from no-user to dual-control, highlighting the challenges of guiding users. Overall, $τ^2$-bench provides a controlled testbed for agents that must both reason effectively and guide user actions.

BibTeX Copy

@misc{barres_2-bench_2025,
	title = {\$τ{\textasciicircum}2\$-{Bench}: {Evaluating} {Conversational} {Agents} in a {Dual}-{Control} {Environment}},
	shorttitle = {\$τ{\textasciicircum}2\$-{Bench}},
	url = {http://arxiv.org/abs/2506.07982},
	doi = {10.48550/arXiv.2506.07982},
	abstract = {Existing benchmarks for conversational AI agents simulate single-control environments, where only the AI agent can use tools to interact with the world, while the user remains a passive information provider. This differs from real-world scenarios like technical support, where users need to actively participate in modifying the state of the (shared) world. In order to address this gap, we introduce \$τ{\textasciicircum}2\$-bench, with four key contributions: 1) A novel Telecom dual-control domain modeled as a Dec-POMDP, where both agent and user make use of tools to act in a shared, dynamic environment that tests both agent coordination and communication, 2) A compositional task generator that programmatically creates diverse, verifiable tasks from atomic components, ensuring domain coverage and controlled complexity, 3) A reliable user simulator tightly coupled with the environment, whose behavior is constrained by tools and observable states, improving simulation fidelity, 4) Fine-grained analysis of agent performance through multiple ablations including separating errors arising from reasoning vs communication/coordination. In particular, our experiments show significant performance drops when agents shift from no-user to dual-control, highlighting the challenges of guiding users. Overall, \$τ{\textasciicircum}2\$-bench provides a controlled testbed for agents that must both reason effectively and guide user actions.},
	urldate = {2026-11-06},
	publisher = {arXiv},
	author = {Barres, Victor and Dong, Honghua and Ray, Soham and Si, Xujie and Narasimhan, Karthik},
	month = jun,
	year = {2025},
	note = {arXiv:2506.07982 [cs.AI]},
	keywords = {Computer Science - Artificial Intelligence, Computer Science - Computation and Language},
}