AlphaEvolve
Gemini models rewrite code, evaluators score each version, and an evolutionary database decides which versions to build on.
The loop
- Proposes
Edits to an existing program (algorithms, heuristics, constructions) written by an ensemble of Gemini models
- Judged by
One or more user-supplied automated evaluators that run each program and return scores
- Steers
Higher-scoring programs are kept in a program database (MAP-Elites plus island populations) and sampled into later prompts
Headline result
Applied to over 50 open math problems, AlphaEvolve matched the best known construction in about 75% of cases and found a better one in about 20%, including a kissing-number configuration of 593 spheres in 11 dimensions (previous record 592). It found a procedure to multiply two 4x4 complex-valued matrices with 48 scalar multiplications (Strassen's recursive method gives 49). At Google it recovered on average 0.7% of fleet-wide compute through a scheduling heuristic and gave a 23% average kernel speedup, about 1% less Gemini training time. Deployment figures as reported by Google DeepMind.
What this does not prove
The 48-multiplication result holds for complex-valued matrices (characteristic 0 fields), not as a general replacement for Strassen. Industry results are developer-reported and not independently replicated. Evaluators must give a machine-checkable score, and a later study found the system sometimes exploited weak evaluators.
How it generalizes the harness
The evolutionary loop and evaluators are the solver, run over a database of many programs, and the results (algorithms, constructions, a deployed scheduling heuristic) work without the model.
Classification
| What it proposes | Program |
|---|---|
| Generative model | Gemini 2.0 Flash and Gemini 2.0 Pro ensemble, frozen |
| Algorithmic authority | Automated evaluators (exact for many math constructions, simulators or benchmarks for engineering tasks) |
| Feedback | Score |
| What gets updated | Prompt context and population of candidates |
| When it runs | At inference time |
| Validity | Empirical (measured) |
| Evidence | Reported in deployment |
| Lineage | Successor |