EoH
Evolution of Heuristics
An LLM writes each heuristic as a short idea in words plus code, and both evolve together under performance-based selection.
The loop
- Proposes
Heuristics as a natural-language 'thought' plus code, made by five LLM prompt operators
- Judged by
Running the code on benchmark instances and measuring fitness
- Steers
Fitness drives selection of parents for the next generation of a fixed-size population
Headline result
On online bin packing, EoH with about 2,000 LLM queries (GPT-3.5-turbo) averaged a 1.18% gap to the lower bound over six test sets, against 2.23% for FunSearch (which the paper says needs around 1 million queries), 4.13% for best fit and 4.51% for first fit. It also reached the best-known tour on three TSPLib instances (pr124, kroA150, u159) and led on Taillard flow-shop sets.
What this does not prove
Main experiments use GPT-3.5-turbo on three classic problems. Comparisons are on the authors' own set-ups, and FunSearch's query count is quoted from its published runs, not re-run.
How it generalizes the harness
The evolutionary loop is the solver over a population, and the output is a heuristic that runs without any model.
Classification
| What it proposes | Heuristic |
|---|---|
| Generative model | Pretrained LLM, frozen (GPT-3.5-turbo in main runs; Gemini Pro, CodeLlama, DeepSeek compared) |
| Algorithmic authority | Benchmark evaluation of the heuristic code |
| Feedback | Score |
| What gets updated | Population of candidates |
| When it runs | Offline, at design time |
| Validity | Empirical (measured) |
| Evidence | Experiments reported |
| Lineage | Original system |