Unified Reinforcement and Imitation Learning for Vision-Language Models

Vision-Language Models (VLMs) have achieved remarkable progress, yet their large scale often renders them impractical for resource-constrained environments. This paper introduces Unified Reinforcement and Imitation Learning (RIL), a novel and efficient training algorithm designed to create powerful, lightweight VLMs. RIL distinctively combines the strengths of reinforcement learning with adversarial imitation learning. This enables smaller student VLMs not only to mimic the sophisticated text generation of large teacher models but also to systematically improve their generative capabilities through reinforcement signals. Key to our imitation framework is an LLM-based discriminator that adeptly distinguishes between student and teacher outputs, complemented by guidance from multiple large teacher VLMs to ensure diverse learning. This unified learning strategy, leveraging both reinforcement and imitation, empowers student models to achieve significant performance gains, making them competitive with leading closed-source VLMs. Extensive experiments on diverse vision-language benchmarks demonstrate that RIL significantly narrows the performance gap with state-of-the-art open-and closed-source VLMs and, in several instances, surpasses them. [Project Page] However, the practical implementation of RIL for VLMs entails several specific challenges. Firstly, relying on continuous discriminator scores (ranging from zero to one) to assess similarity to large VLM outputs can introduce ambiguity into the learning signal. To ensure a clearer and more decisive reward, and drawing inspiration from prior works that binarize answer rewards [7, 43] , we convert the discriminator's similarity score into a binary value. Secondly, the discriminator, by design, focuses on stylistic similarity and does not inherently verify factual correctness against ground truth