Reinforced Curriculum Pre-Alignment for Domain-Adaptive VLMs

TL;DR

This paper introduces RCPA, a curriculum-based post-training method for adapting vision-language models to specialized domains while maintaining their general capabilities, addressing catastrophic forgetting and optimization challenges.

Contribution

The paper proposes Reinforced Curriculum Pre-Alignment (RCPA), a novel staged adaptation approach that safely exposes models to new domains and preserves their general abilities.

Findings

RCPA outperforms existing methods in domain-specific tasks.

It effectively balances domain adaptation with general capability retention.

Experimental results validate RCPA's robustness across multiple benchmarks.

Abstract

Vision-Language Models (VLMs) demonstrate remarkable general-purpose capabilities but often fall short in specialized domains such as medical imaging or geometric problem-solving. Supervised Fine-Tuning (SFT) can enhance performance within a target domain, but it typically causes catastrophic forgetting, limiting its generalization. The central challenge, therefore, is to adapt VLMs to new domains while preserving their general-purpose capabilities. Continual pretraining is effective for expanding knowledge in Large Language Models (LLMs), but it is less feasible for VLMs due to prohibitive computational costs and the unavailability of pretraining data for most open-source models. This necessitates efficient post-training adaptation methods. Reinforcement learning (RL)-based approaches such as Group Relative Policy Optimization (GRPO) have shown promise in preserving general abilities, yet they often fail in domain adaptation scenarios where the model initially lacks sufficient domain knowledge, leading to optimization collapse. To bridge this gap, we propose Reinforced Curriculum Pre-Alignment (RCPA), a novel post-training paradigm that introduces a curriculum-aware progressive modulation mechanism. In the early phase, RCPA applies partial output constraints to safely expose the model to new domain concepts. As the model's domain familiarity increases, training gradually transitions to full generation optimization, refining responses and aligning them with domain-specific preferences. This staged adaptation balances domain knowledge acquisition with the preservation of general multimodal capabilities. Extensive experiments across specialized domains and general benchmarks validate the effectiveness of RCPA, establishing a practical pathway toward building high-performing and domain-adaptive VLMs.