titleDo LLM Agents Still Need to Be Told How to Self-Improve?
authorsShobhita Sundaram, Cansu Sancaktar, Phillip Isola
date2026-11-17
venuearXiv (preprint, not peer reviewed)
idarXiv:2611.23219
affiliationMassachusetts Institute of Technology
upvotes5 (Hugging Face Daily Papers, 2026-11-19)
linksabs pdf hf code
cite
relatedEvoArena: Tracking Memory Evolution for Robust LLM Agents in Dynamic Environments
From Activation to Causality: Discovery of Causal Visual Representations in the Human Brain
Learning from Teacher Continuations at Student States
DySENSE: Dynamic Spatiotemporal ENergy Synthesis and Emulation with Satellite Imagery
Base Models Can Reason By Taking a Cue From Training Data
abstractLLM agents increasingly drive their own learning. Self-improvement methods delegate various components of model optimization to the models themselves. As models become more capable, this motivates a basic question: do these increasingly capable models still need us to specify how they should self-improve? In this paper we study an autonomous paradigm in which, given basic primitives, resources, and a general goal for self-improvement, agents construct their own self-learning algorithm for context-based optimization. Using the setting of gameplay as a testbed, a frozen LLM agent is given a compute budget and tools for collecting experience and updating text documents, and allowed to freely experiment without an externally imposed loop or access to real-world data. We find that sufficiently capable models do self-improve under these conditions, demonstrating strategies that align with classic cognitive frameworks for self-regulated learning and constructs from machine learning literature on self-generated curricula. Furthermore, we study the extent to which the learned text artifacts transfer to external benchmarks, observing improvement across different game horizons, mechanics, and 3D navigation.

bibtex

@misc{sundaram2026do,
  title={Do LLM Agents Still Need to Be Told How to Self-Improve?},
  author={Shobhita Sundaram and Cansu Sancaktar and Phillip Isola},
  year={2026},
  eprint={2611.23219},
  archivePrefix={arXiv},
  url={https://arxiv.org/abs/2611.23219}
}