TY - JOUR
AU - 
AB - Target: picture Communication between human and mobile Should I stay on this floor or go down the agents is getting increasingly important as stairs? such agents are widely deployed in our daily Yeah, go down the stairs. And then I think lives. Vision-and-Dialogue Navigation is one you turn right but I can't really tell. of the tasks that evaluate the agent’s ability to Should I go left toward the door, or right interact with humans for assistance and navi- around the corner? gate based on natural language responses. In this paper, we explore the Navigation from Di- Go to the left, through that door. alogue History (NDH) task, which is based on the Cooperative Vision-and-Dialogue Nav- Figure 1: One example in the CVDN dataset. Given igation (CVDN) dataset, and present a state- target information, dialogues in blue text and red text of-the-art model which is built upon Vision- sequentially, the human navigates the green path, blue Language transformers. However, despite path, and red path accordingly. achieving competitive performance, we find that the agent in the NDH task is not evaluated most effective ways of communication due to its appropriately by the primary metric – Goal flexibility. Therefore, many efforts 
TI - NDH-Full: Learning and Evaluating Navigational Agents on Full-Length Dialogue
JF - Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
DO - 10.18653/v1/2021.emnlp-main.518
DA - 2021-01-01
UR - https://www.deepdyve.com/lp/unpaywall/ndh-full-learning-and-evaluating-navigational-agents-on-full-length-hdfvId2NA5
DP - DeepDyve
ER -