Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.18743 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918397304897536 |
|---|---|
| author | Zhou, Huichi Guo, Siyuan Liu, Anjie Yu, Zhongwei Gong, Ziqin Zhao, Bowen Chen, Zhixun Zhang, Menglong Chen, Yihang Li, Jinsong Yang, Runyu Liu, Qiangbin Yu, Xinlei Zhou, Jianmin Wang, Na Sun, Chunyang Wang, Jun |
| author_facet | Zhou, Huichi Guo, Siyuan Liu, Anjie Yu, Zhongwei Gong, Ziqin Zhao, Bowen Chen, Zhixun Zhang, Menglong Chen, Yihang Li, Jinsong Yang, Runyu Liu, Qiangbin Yu, Xinlei Zhou, Jianmin Wang, Na Sun, Chunyang Wang, Jun |
| contents | We introduce \emph{Memento-Skills}, a generalist, continually-learnable LLM agent system that functions as an \emph{agent-designing agent}: it autonomously constructs, adapts, and improves task-specific agents through experience. The system is built on a memory-based reinforcement learning framework with \emph{stateful prompts}, where reusable skills (stored as structured markdown files) serve as persistent, evolving memory. These skills encode both behaviour and context, enabling the agent to carry forward knowledge across interactions.
Starting from simple elementary skills (like Web search and terminal operations), the agent continually improves via the \emph{Read--Write Reflective Learning} mechanism introduced in \emph{Memento~2}~\cite{wang2025memento2}. In the \emph{read} phase, a behaviour-trainable skill router selects the most relevant skill conditioned on the current stateful prompt; in the \emph{write} phase, the agent updates and expands its skill library based on new experience. This closed-loop design enables \emph{continual learning without updating LLM parameters}, as all adaptation is realised through the evolution of externalised skills and prompts.
Unlike prior approaches that rely on human-designed agents, Memento-Skills enables a generalist agent to \emph{design agents end-to-end} for new tasks. Through iterative skill generation and refinement, the system progressively improves its own capabilities. Experiments on the \emph{General AI Assistants} benchmark and \emph{Humanity's Last Exam} demonstrate sustained gains, achieving 26.2\% and 116.2\% relative improvements in overall accuracy, respectively. Code is available at https://github.com/Memento-Teams/Memento-Skills. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_18743 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Memento-Skills: Let Agents Design Agents Zhou, Huichi Guo, Siyuan Liu, Anjie Yu, Zhongwei Gong, Ziqin Zhao, Bowen Chen, Zhixun Zhang, Menglong Chen, Yihang Li, Jinsong Yang, Runyu Liu, Qiangbin Yu, Xinlei Zhou, Jianmin Wang, Na Sun, Chunyang Wang, Jun Artificial Intelligence Computation and Language Machine Learning We introduce \emph{Memento-Skills}, a generalist, continually-learnable LLM agent system that functions as an \emph{agent-designing agent}: it autonomously constructs, adapts, and improves task-specific agents through experience. The system is built on a memory-based reinforcement learning framework with \emph{stateful prompts}, where reusable skills (stored as structured markdown files) serve as persistent, evolving memory. These skills encode both behaviour and context, enabling the agent to carry forward knowledge across interactions. Starting from simple elementary skills (like Web search and terminal operations), the agent continually improves via the \emph{Read--Write Reflective Learning} mechanism introduced in \emph{Memento~2}~\cite{wang2025memento2}. In the \emph{read} phase, a behaviour-trainable skill router selects the most relevant skill conditioned on the current stateful prompt; in the \emph{write} phase, the agent updates and expands its skill library based on new experience. This closed-loop design enables \emph{continual learning without updating LLM parameters}, as all adaptation is realised through the evolution of externalised skills and prompts. Unlike prior approaches that rely on human-designed agents, Memento-Skills enables a generalist agent to \emph{design agents end-to-end} for new tasks. Through iterative skill generation and refinement, the system progressively improves its own capabilities. Experiments on the \emph{General AI Assistants} benchmark and \emph{Humanity's Last Exam} demonstrate sustained gains, achieving 26.2\% and 116.2\% relative improvements in overall accuracy, respectively. Code is available at https://github.com/Memento-Teams/Memento-Skills. |
| title | Memento-Skills: Let Agents Design Agents |
| topic | Artificial Intelligence Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2603.18743 |