Agent KB : Leveraging Cross-Domain Experience for Agentic Problem Solving
Xiangru Tang O, Tianrui Qin O, Tianhao Peng O, Ziyang Zhou, Daniel Shao, Tingting Du · 2025
Read Agent KB : Leveraging Cross-Domain Experience for Agentic Problem Solving by Xiangru Tang O, Tianrui Qin O, Tianhao Peng O, Ziyang Zhou, Daniel Shao, Tingting Du (2025) 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-2507-06229.txt.
Description
Xiangru Tang and co-authors present Agent KB in this 2025 paper, a shared memory system for AI agents. Agent frameworks today work alone, so each one repeats the same errors and solves the same problems twice. Past attempts and fixes stay trapped inside one framework. Agent KB gathers these traces into one knowledge base that any framework can query.
At run time, the system helps in two stages. First, plans from past work seed the agent with cross-domain workflows. Then, feedback adds targeted fixes when the agent gets stuck. A gate blocks retrieved tips that would clash with the agent's own reasoning. Tests on four benchmarks showed clear gains for several frameworks and model families. The authors present shared memory as a base for collective agent intelligence.
Catalog record
| CREATOR | Xiangru Tang O, Tianrui Qin O, Tianhao Peng O, Ziyang Zhou, Daniel Shao, Tingting Du |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 147 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-06229