Akiba et al.: evolutionary search over parameter and dataflow model merges

What changed

Optimized model-merging recipes in parameter and inference-dataflow spaces using evolutionary search, combining capabilities of existing models without retraining each from scratch.

Scientific scope

Optimized model-merging recipes in parameter and inference-dataflow spaces using evolutionary search, combining capabilities of existing models without retraining each from scratch. Model merging, pretrained weights and evolutionary optimization are inherited. Search requires evaluation data and compute; no-additional-training is not zero total cost or universally compatible weights.2025 journal revision is not backdated into original2024 results.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Model merging, pretrained weights and evolutionary optimization are inherited. Search requires evaluation data and compute; no-additional-training is not zero total cost or universally compatible weights.2025 journal revision is not backdated into original2024 results.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Evolutionary Optimization of Model Merging Recipes
• Akiba et al.: evolutionary search over parameter and dataflow model merges — author/institutional project

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction