ContextEvolve: Multi-Agent Context Compression for Systems Code Optimization
Hongyuan Su, Yu Zheng, Yong Li · 2026
Read ContextEvolve: Multi-Agent Context Compression for Systems Code Optimization by Hongyuan Su, Yu Zheng, Yong Li (2026) 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-2602-02597.txt.
Description
Hongyuan Su and 2 co-authors compress the working context of language models that optimize systems code, in a 2026 arXiv paper. Models now discover performance-critical algorithms, but meeting systems correctness and performance demands needs iterative search. Test-time reinforcement learning searches efficiently yet needs setting updates impossible through API-only access. Training-free evolutionary methods waste context and wander undirected.
ContextEvolve splits tuning context among three agents. A summarizer condenses semantic state by turning code into language. A navigator distills tuning direction from trajectory analysis. A sampler curates experience through prioritized exemplar retrieval. Together they mirror reinforcement learning's state representation, policy gradient, and experience replay inside a textual latent space. On the ADRS benchmark the system beat state-of-the-art rivals by about a third while cutting token consumption by about twenty-nine percent.
Catalog record
| CREATOR | Hongyuan Su, Yu Zheng, Yong Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02597