Tuning Large Multimodal Models for Videos using Reinforcement Learning from AI Feedback
This work addresses the problem of deficient multimodal instruction-tune data for video-text alignment, offering a potential improvement for researchers and developers in video understanding and generation, though it appears incremental as it builds on existing RLAIF concepts applied to a new domain.
The paper tackles the challenge of aligning video and text modalities in large multimodal models by introducing a novel Reinforcement Learning from AI Feedback (RLAIF) strategy, which uses self-preference feedback and context-aware reward modeling to outperform existing methods like Supervised Fine-Tuning on diverse video benchmarks.
Recent advancements in large language models have influenced the development of video large multimodal models (VLMMs). The previous approaches for VLMMs involved Supervised Fine-Tuning (SFT) with instruction-tuned datasets, integrating LLM with visual encoders, and adding additional learnable modules. Video and text multimodal alignment remains challenging, primarily due to the deficient volume and quality of multimodal instruction-tune data compared to text-only data. We present a novel alignment strategy that employs multimodal AI system to oversee itself called Reinforcement Learning from AI Feedback (RLAIF), providing self-preference feedback to refine itself and facilitating the alignment of video and text modalities. In specific, we propose context-aware reward modeling by providing detailed video descriptions as context during the generation of preference feedback in order to enrich the understanding of video content. Demonstrating enhanced performance across diverse video benchmarks, our multimodal RLAIF approach, VLM-RLAIF, outperforms existing approaches, including the SFT model. We commit to open-sourcing our code, models, and datasets to foster further research in this area.