Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.01966 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914301379346432 |
|---|---|
| author | Yu, Hongzhuo Zhu, Fei Xie, Guo-Sen Shao, Ling |
| author_facet | Yu, Hongzhuo Zhu, Fei Xie, Guo-Sen Shao, Ling |
| contents | While large language model (LLM) agents have demonstrated impressive problem-solving capabilities, they typically operate as static systems, lacking the ability to evolve through lifelong interaction. Existing attempts to bridge this gap primarily rely on retrieving successful past trajectories as demonstrations. However, this paradigm faces two critical limitations. First, by focusing solely on success, agents overlook the rich pedagogical value embedded in failed attempts, preventing them from identifying and avoiding recurrent pitfalls. Second, continually accumulating textual experiences not only increases the time consumption during retrieval but also inevitably introduces noise and exhausts the largest context window of current LLMs. To address these challenges, we propose a novel self-evolving framework for LLM agents that introduces a complementary evolution mechanism: First, a contrastive reflection strategy is introduced to explicitly summarize error-prone patterns and capture reusable insights. Second, we propose a self-consolidation mechanism that distills non-parametric textual experience into compact learnable parameters. This enables the agent to internalize extensive historical experience directly into its latent space. Extensive experiments demonstrate the advantages of our method in long-term agent evolution. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01966 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Self-Consolidation for Self-Evolving Agents Yu, Hongzhuo Zhu, Fei Xie, Guo-Sen Shao, Ling Machine Learning While large language model (LLM) agents have demonstrated impressive problem-solving capabilities, they typically operate as static systems, lacking the ability to evolve through lifelong interaction. Existing attempts to bridge this gap primarily rely on retrieving successful past trajectories as demonstrations. However, this paradigm faces two critical limitations. First, by focusing solely on success, agents overlook the rich pedagogical value embedded in failed attempts, preventing them from identifying and avoiding recurrent pitfalls. Second, continually accumulating textual experiences not only increases the time consumption during retrieval but also inevitably introduces noise and exhausts the largest context window of current LLMs. To address these challenges, we propose a novel self-evolving framework for LLM agents that introduces a complementary evolution mechanism: First, a contrastive reflection strategy is introduced to explicitly summarize error-prone patterns and capture reusable insights. Second, we propose a self-consolidation mechanism that distills non-parametric textual experience into compact learnable parameters. This enables the agent to internalize extensive historical experience directly into its latent space. Extensive experiments demonstrate the advantages of our method in long-term agent evolution. |
| title | Self-Consolidation for Self-Evolving Agents |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2602.01966 |