MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866918333243195392 |
|---|---|
| author | Zhang, Shengtao Wang, Jiaqian Zhou, Ruiwen Liao, Junwei Feng, Yuchen Li, Zhuo Zheng, Yujie Zhang, Weinan Wen, Ying Li, Zhiyu Xiong, Feiyu Qi, Yutao Tang, Bo Wen, Muning |
| author_facet | Zhang, Shengtao Wang, Jiaqian Zhou, Ruiwen Liao, Junwei Feng, Yuchen Li, Zhuo Zheng, Yujie Zhang, Weinan Wen, Ying Li, Zhiyu Xiong, Feiyu Qi, Yutao Tang, Bo Wen, Muning |
| contents | The hallmark of human intelligence is the self-evolving ability to master new skills by learning from past experiences. However, current AI agents struggle to emulate this self-evolution: fine-tuning is computationally expensive and prone to catastrophic forgetting, while existing memory-based methods rely on passive semantic matching that often retrieves noise. To address these challenges, we propose MemRL, a non-parametric approach that evolves via reinforcement learning on episodic memory. By decoupling stable reasoning from plastic memory, MemRL employs a Two-Phase Retrieval mechanism to filter noise and identify high-utility strategies through environmental feedback. Extensive experiments on HLE, BigCodeBench, ALFWorld, and Lifelong Agent Bench demonstrate that MemRL significantly outperforms state-of-the-art baselines, confirming that MemRL effectively reconciles the stability-plasticity dilemma, enabling continuous runtime improvement without weight updates. Code is available at https://github.com/MemTensor/MemRL. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_03192 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory Zhang, Shengtao Wang, Jiaqian Zhou, Ruiwen Liao, Junwei Feng, Yuchen Li, Zhuo Zheng, Yujie Zhang, Weinan Wen, Ying Li, Zhiyu Xiong, Feiyu Qi, Yutao Tang, Bo Wen, Muning Computation and Language The hallmark of human intelligence is the self-evolving ability to master new skills by learning from past experiences. However, current AI agents struggle to emulate this self-evolution: fine-tuning is computationally expensive and prone to catastrophic forgetting, while existing memory-based methods rely on passive semantic matching that often retrieves noise. To address these challenges, we propose MemRL, a non-parametric approach that evolves via reinforcement learning on episodic memory. By decoupling stable reasoning from plastic memory, MemRL employs a Two-Phase Retrieval mechanism to filter noise and identify high-utility strategies through environmental feedback. Extensive experiments on HLE, BigCodeBench, ALFWorld, and Lifelong Agent Bench demonstrate that MemRL significantly outperforms state-of-the-art baselines, confirming that MemRL effectively reconciles the stability-plasticity dilemma, enabling continuous runtime improvement without weight updates. Code is available at https://github.com/MemTensor/MemRL. |
| title | MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2601.03192 |