analiza aiqbench
This research introduces "Evolutionary Soups," a novel Mixture-of-Experts framework that uses evolutionary algorithms to dynamically optimize gating networks for multi-objective LLM alignment at inference time. By leveraging hidden-state representations, the method achieves significant improvements in utility metrics across noisy datasets without requiring model retraining. This advancement offers a promising pathway for enhancing controllable generation in complex scenarios where competing objectives must be balanced dynamically.
Evolutionary Soups: Evolving Mixture-of-Experts for Multi-Objective LLM Alignment
arXiv:2608.29978v1 Announce Type: new Abstract: Large language models are increasingly required to generate responses that satisfy multiple competing objectives. Since optimal trade-offs depend on both user preferences and input prompts, controllable multi-objective generation must dynamically adapt models at inference time without retraining. To address this, we propose Evolutionary Soups, a mixture-of-experts framework for fine-grained generation control, with gating networks trained via an evolutionary algorithm. The per-layer gating networks dynamically produce expert-merging coefficients from hidden-state representations, while the evolutionary algorithm incorporates greedy hypervolume contribution for effective evolution of these gating networks, achieving consistent improvements on large and noisy training datasets and broader coverage of the non-convex Pareto front. Experiments across three tasks demonstrate the effectiveness of Evolutionary Soups over baselines: it achieves the best hypervolume, linear utility, and Tchebyshev utility (~20% improvement) among controllable methods on all tasks.
Abstract:Large language models are increasingly required to generate responses that satisfy multiple competing objectives. Since optimal trade-offs depend on both user preferences and input prompts, controllable multi-objective generation must dynamically adapt models at inference time without retraining. To address this, we propose Evolutionary Soups, a mixture-of-experts framework for fine-grained generation control, with gating networks trained via an evolutionary algorithm. The per-layer gating networks dynamically produce expert-merging coefficients from hidden-state representations, while the evolutionary algorithm incorporates greedy hypervolume contribution for effective evolution of these gating networks, achieving consistent improvements on large and noisy training datasets and broader coverage of the non-convex Pareto front. Experiments across three tasks demonstrate the effectiveness of Evolutionary Soups over baselines: it achieves the best hypervolume, linear utility, and Tchebyshev utility (~20% improvement) among controllable methods on all tasks.
Computation and Language (cs.CL); Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE)
arXiv:2608.29978 [cs.CL]
(or arXiv:2608.29978v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2608.29978 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 235 cuvinte
Subiecte conexe
Humanity’s Last Invention — Richard Socher of Recursive
Latent Space · 14 sept. 2026
OneBid: A Unified Auto-Bidding Foundation Model for Diverse oCPX Advertising Scenarios
arXiv · 22 sept. 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · 13 oct. 2026
[AINews] Agents disprove Polysemanticity; greatly improve Interpretability & Alignment
Latent Space · 16 oct. 2026