![[2602.13792] StackingNet: Collective Inference Across Independent AI Foundation Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13792] StackingNet: Collective Inference Across Independent AI Foundation Models
Summary
StackingNet introduces a meta-ensemble framework that enhances the coordination of independent AI foundation models, improving accuracy, fairness, and reliability in various tasks.
Why It Matters
As AI systems increasingly rely on large foundation models, the ability to integrate their diverse strengths is crucial for developing trustworthy and robust intelligent systems. StackingNet offers a novel approach to leverage these models collectively, which could lead to significant advancements in AI applications across multiple domains.
Key Takeaways
- StackingNet improves model accuracy and reduces bias by coordinating independent AI models.
- The framework operates without needing access to internal parameters or training data.
- It enhances reliability ranking and identifies models that may degrade performance.
- StackingNet promotes collaboration among diverse models, turning inconsistency into strength.
- The approach suggests that future AI advancements may come from cooperation rather than just larger models.
Computer Science > Artificial Intelligence arXiv:2602.13792 (cs) [Submitted on 14 Feb 2026] Title:StackingNet: Collective Inference Across Independent AI Foundation Models Authors:Siyang Li, Chenhao Liu, Dongrui Wu, Zhigang Zeng, Lieyun Ding View a PDF of the paper titled StackingNet: Collective Inference Across Independent AI Foundation Models, by Siyang Li and 4 other authors View PDF HTML (experimental) Abstract:Artificial intelligence built on large foundation models has transformed language understanding, vision and reasoning, yet these systems remain isolated and cannot readily share their capabilities. Integrating the complementary strengths of such independent foundation models is essential for building trustworthy intelligent systems. Despite rapid progress in individual model design, there is no established approach for coordinating such black-box heterogeneous models. Here we show that coordination can be achieved through a meta-ensemble framework termed StackingNet, which draws on principles of collective intelligence to combine model predictions during inference. StackingNet improves accuracy, reduces bias, enables reliability ranking, and identifies or prunes models that degrade performance, all operating without access to internal parameters or training data. Across tasks involving language comprehension, visual estimation, and academic paper rating, StackingNet consistently improves accuracy, robustness, and fairness, compared with individual models and cla...