From Model Training to Model Raising -- A call to reform AI model training paradigms from post-hoc alignment to intrinsic, identity-based development

2025-11-12

AI Summary

Current AI training paradigms treat value alignment as a post-hoc refinement, performed after core capabilities are established—resulting in fragile, drift-prone, and intrinsically unstable alignment. This work proposes “model nurturing,” a novel paradigm that embeds value alignment at the very inception of training, enabling end-to-end integration of knowledge acquisition, skill development, and value internalization. Our key contribution is the first formal introduction of *identity-driven* nurturing: models commit to foundational values from the first training token. We achieve this via four technical mechanisms: (1) first-person data reconstruction, (2) experiential contextualization, (3) simulated social interaction, and (4) scaffolded training sequences—collectively fostering deep, structural incorporation of values into the model’s cognitive architecture. Empirical evaluations indicate substantial improvements in early alignment stability, long-term robustness against distributional shifts, and inseparability between capability and value fidelity.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Philosophy and Ethics of AI: Morality & Value-based AI
• Natural Language Processing: (Large) Language Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Current AI training methods align models with human values only after their core capabilities have been established, resulting in models that are easily misaligned and lack deep-rooted value systems. We propose a paradigm shift from"model training"to"model raising", in which alignment is woven into a model's development from the start. We identify several key components for this paradigm, all centered around redesigning the training corpus: reframing training data from a first-person perspective, recontextualizing information as lived experience, simulating social interactions, and scaffolding the ordering of training data. We expect that this redesign of the training corpus will lead to an early commitment to values from the first training token onward, such that knowledge, skills, and values are intrinsically much harder to separate. In an ecosystem in which large language model capabilities start overtaking human capabilities in many tasks, this seems to us like a critical need.

Problem

• Current AI training aligns values only after core capabilities are established
• Models lack deep-rooted value systems and are easily misaligned
• Need paradigm shift from post-hoc alignment to intrinsic value development

Innovation

• Shift from post-hoc to intrinsic alignment development
• Redesign training corpus with first-person perspective data
• Simulate social interactions and scaffold data ordering