# Revisiting EmbodiedQA: A Simple Baseline and Beyond

## TL;DR

This paper presents a simple baseline for EmbodiedQA, introduces a practical setting allowing model adaptation to new environments, and demonstrates competitive results and improved navigation performance.

## Contribution

It proposes a straightforward baseline and a new adaptable setting for EmbodiedQA, enhancing generalization and navigation success.

## Key findings

- The baseline achieves competitive results on EmbodiedQA v1.
- Adapting models in new environments improves navigation performance.
- The new setting offers a practical approach for real-world applications.

## Abstract

In Embodied Question Answering (EmbodiedQA), an agent interacts with an environment to gather necessary information for answering user questions. Existing works have laid a solid foundation towards solving this interesting problem. But the current performance, especially in navigation, suggests that EmbodiedQA might be too challenging for the contemporary approaches. In this paper, we empirically study this problem and introduce 1) a simple yet effective baseline that achieves promising performance; 2) an easier and practical setting for EmbodiedQA where an agent has a chance to adapt the trained model to a new environment before it actually answers users questions. In this new setting, we randomly place a few objects in new environments, and upgrade the agent policy by a distillation network to retain the generalization ability from the trained model. On the EmbodiedQA v1 benchmark, under the standard setting, our simple baseline achieves very competitive results to the-state-of-the-art; in the new setting, we found the introduced small change in settings yields a notable gain in navigation.