EvidenceDemonstrated real incidentv1.22.1
In plain English
This page shows what kind of support exists for each claim: real systems, experiments, early evidence, architectural reasoning, open questions, or speculative scenarios.
- Why this matters: AI risk can come from the whole arrangement, not one obvious model.
- What to look for: data, memory, routes, adapters, tools, evaluators, updates, and rollback paths.
- Technical version below: the expert terminology remains available and is linked through the glossary.
Evolutionary optimization of model merging recipes
Evidence card
- Claim
- Automated search can discover high-performing combinations of open-source models and data-flow recipes.
- Evidence level
- Demonstrated
- Source
- https://www.nature.com/articles/s42256-024-00975-8
- Publication date
- 2025-01-28
- Authors or institution
- Takuya Akiba, Makoto Shing, Yujin Tang, Qi Sun, David Ha
- System tested
- Evolutionary search over model merging recipes in parameter space and data-flow space.
- Limitations
- Utility-focused model development; not a system-safety guarantee for adaptive ecologies.
- What the evidence does show
- Automated search can discover high-performing combinations of open-source models and data-flow recipes.
- What the evidence does not show
- That automated model merging preserves safety properties under all derived compositions.
- Date last reviewed in UTC
- 2026-06-26T00:00:00Z
Site use
This source supports Cognivirus.com pages related to evolutionary model mergingCombining model weights or adapter deltas into one artifact. Open glossary definition, model composition, automated model development. Its role is bounded by the limitations listed above.