Self-Evolving Agents via Likelihood-Guided Tool-Space Optimization

Research overview

LOTS scores tool outputs through changes in answer likelihood and consolidates accumulated experience into reusable task-specific tool spaces while keeping model weights fixed.

Metadata checked against arXiv and Google Scholar on October 6, 2026. This overview summarizes the work; consult the linked paper for results and limitations.

Citation

@misc{lots2026,
  title={Self-Evolving Agents via Likelihood-Guided Tool-Space Optimization},
  author={Zhang, Xuanqi and Jin, Ruinan and Yang, Running and Zhang, Yuxuan and Chen, Minghui and Deng, Wenlong and Li, Xiaoxiao},
  year={2026},
  eprint={2609.34151},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2609.34151}
}