A Self-Improving Coding Agent
Maxime Robeyns, Martin Szummer, Laurence aitchison · 2025
Read A Self-Improving Coding Agent by Maxime Robeyns, Martin Szummer, Laurence aitchison (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-15228.txt.
Description
Maxime Robeyns, Martin Szummer, and Laurence Aitchison show a coding agent that rewrites itself to get better. Their 2025 paper builds an agent system: code that calls a large language model and hands it basic coding tools. The agent turns those tools on its own code, editing its design in light of past runs. No gradient training drives the gains, only reflection and code change.
Scores climbed steeply on a slice of a software repair benchmark, with further gains on a live coding benchmark and on synthetic agent tests. The authors frame the loop as a data-light path to open-ended agent design, where systems improve their own scaffolding over time.
Catalog record
| CREATOR | Maxime Robeyns, Martin Szummer, Laurence aitchison |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-15228