使用持续学习改进多模态大语言模型
Improving Multimodal Large Language Models Using Continual Learning
AI总结：
本研究将视觉模型集成至LLM视为持续学习问题，评估五种方法缓解遗忘，提出在保持多模态准确率的同时将语言性能退化降低高达15%的技术。
摘要：
生成式大语言模型（LLMs）展现出令人印象深刻的能力，通过将预训练视觉模型集成到原始LLM中创建多模态LLM（MLLM），可以进一步增强这种能力。然而，与原始LLM相比，这种集成通常会显著降低自然语言理解和生成任务的性能。本研究使用LLaVA MLLM调查此问题，将该集成视为持续学习问题。我们评估了五种持续学习方法以缓解遗忘，并确定了一种在最小化语言性能损失的同时增强视觉理解的技术。我们的方法在LLaVA方案上减少了高达15%的语言性能退化，同时保持了高多模态准确率。我们还通过在一系列视觉-语言任务上进行持续学习，证明了我们方法的鲁棒性，在获取新多模态能力的同时有效保留了语言技能。项目网页：https://shikhar-srivastava.github.io/cl-for-improving-mllms
英文摘要：
Generative large language models (LLMs) exhibit impressive capabilities, which can be further augmented by integrating a pre-trained vision model into the original LLM to create a multimodal LLM (MLLM). However, this integration often significantly decreases performance on natural language understanding and generation tasks, compared to the original LLM. This study investigates this issue using the LLaVA MLLM, treating the integration as a continual learning problem. We evaluate five continual learning methods to mitigate forgetting and identify a technique that enhances visual understanding while minimizing linguistic performance loss. Our approach reduces linguistic performance degradation by up to 15% over the LLaVA recipe, while maintaining high multimodal accuracy. We also demonstrate the robustness of our method through continual learning on a sequence of vision-language tasks, effectively preserving linguistic skills while acquiring new multimodal capabilities. Project webpage: https://shikhar-srivastava.github.io/cl-for-improving-mllms
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
