ToolSelf : Unifying Task Execution and Self-Reconfiguration via Tool-Driven Intrinsic Adaptation

Jingqi Zhou, Sheng Wang, DeZhao Deng, Junwen Lu, Junwei Su, Qintong Li, Jiahui Gao, Hao Wu, Jiyue Jiang, Lingpeng Kong, Chuan Wu · 2026

Read ToolSelf : Unifying Task Execution and Self-Reconfiguration via Tool-Driven Intrinsic Adaptation by Jingqi Zhou, Sheng Wang, DeZhao Deng, Junwen Lu, Junwei Su, Qintong Li, Jiahui Gao, Hao Wu, Jiyue Jiang, Lingpeng Kong, Chuan Wu (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-2602-07883.txt.

Description

AI agents usually run with configurations fixed before the task begins, forcing a choice between specialized strength and broad generality. Jingqi Zhou and eleven co-authors propose ToolSelf, a paradigm in which the executing agent reconfigures itself during the task, treating configuration updates as ordinary tool calls within one policy.

As work progresses, the agent can revise sub-goals, strategies, toolboxes, context, and context-management modes in response to feedback, unifying execution and adaptation instead of separating them. A two-stage configuration-aware training scheme teaches this behavior. The paper addresses a real tension in long-horizon agents: plans made before seeing the task rarely survive contact with it. ToolSelf lets the agent adapt without handing control to a separate optimizer.

Catalog record

CREATORJingqi Zhou, Sheng Wang, DeZhao Deng, Junwen Lu, Junwei Su, Qintong Li, Jiahui Gao, Hao Wu, Jiyue Jiang, Lingpeng Kong, Chuan Wu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE101 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-07883

Browse the full library · About the Tower of Babel