OR-Agent: Bridging Evolutionary Search and Structured Research for Automated Algorithm Discovery
Qi Liu, Ruochen Hao*, Can Li, Wanjing Ma · 2026
Read OR-Agent: Bridging Evolutionary Search and Structured Research for Automated Algorithm Discovery by Qi Liu, Ruochen Hao*, Can Li, Wanjing Ma (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-2602-13769.txt.
Description
Automating heuristic design in lab-driven domains needs more than step-by-step mutation of solution algorithms. Current language-model evolution methods rely on random mutation loops without long-term planning or learning from past failures, causing inefficient search and redundant trials. OR-Agent organizes heuristic search as a tree-based workflow modeling branching hypothesis generation and steady backtracking, within a multi-agent research framework for tuning problems with rich lab environments.
A layered, tuning-inspired review system adds adaptive learning: short-term reviews act as verbal gradients, long-term reviews as verbal momentum, and memory compression as semantic weight decay, together steering research work. Experiments on classic combo problems such as traveling salesman, vehicle routing, and bin packing, plus simulated team driving, show OR-Agent beating strong evolution search rivals. Qi Liu leads the work with three co-authors.
Catalog record
| CREATOR | Qi Liu, Ruochen Hao*, Can Li, Wanjing Ma |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 99 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-13769