MLLM as Retriever: Interactively Learning Multimodal Retrieval for Embodied Agents
Junpeng Yue, Xinru Xu, Börje F. Karlsson, Zongqing Lu
2024-10-13
Summary
This paper introduces MLLM as ReTriever (MART), a new method that helps AI agents learn to retrieve useful information from various types of data (like text and images) to perform tasks more effectively.
What's the problem?
Current methods for helping AI agents find and use information often only look at surface similarities in the data, such as matching words or images. This approach can miss important details that are necessary for completing specific tasks, making it hard for agents to perform well in complex environments.
What's the solution?
To improve this, the authors developed MART, which uses interaction data to fine-tune a large language model (LLM) so that it can better assess the effectiveness of different paths or actions (called trajectories) for completing tasks. They also introduced a technique called Trajectory Abstraction, which simplifies these trajectories by summarizing them while keeping the important information. This allows the AI agents to understand key milestones more easily. The results showed that MART significantly improved the success rates of AI agents in new environments compared to existing methods.
Why it matters?
This research is important because it presents a new way for AI agents to learn from their experiences and improve their performance in real-world tasks. By enhancing how these agents retrieve and process multimodal information, MART can lead to better outcomes in applications like robotics, virtual assistants, and gaming, where understanding complex environments is crucial.
Abstract
MLLM agents demonstrate potential for complex embodied tasks by retrieving multimodal task-relevant trajectory data. However, current retrieval methods primarily focus on surface-level similarities of textual or visual cues in trajectories, neglecting their effectiveness for the specific task at hand. To address this issue, we propose a novel method, MLLM as ReTriever (MART), which enhances the performance of embodied agents by utilizing interaction data to fine-tune an MLLM retriever based on preference learning, such that the retriever fully considers the effectiveness of trajectories and prioritize them for unseen tasks. We also introduce Trajectory Abstraction, a mechanism that leverages MLLMs' summarization capabilities to represent trajectories with fewer tokens while preserving key information, enabling agents to better comprehend milestones in the trajectory. Experimental results across various environments demonstrate our method significantly improves task success rates in unseen scenes compared to baseline methods. This work presents a new paradigm for multimodal retrieval in embodied agents, by fine-tuning a general-purpose MLLM as the retriever to assess trajectory effectiveness. All benchmark task sets and simulator code modifications for action and observation spaces will be released.