AI systems embodied in the physical world face a fundamental challenge of partial observability; operating with only a limited view and knowledge of the environment. This creates challenges when AI systems try to reason about language and its relationship with the environment: objects referred to through language (e.g. giving many instructions) are not immediately visible. Actions by the AI system may be required to bring these objects in view. A good benchmark to study this challenge is Dynamic Referring Expression Recognition (dRER) task, where the goal is to find a target location by dynamically adjusting the field of view (FoV) in a partially observed 360 scenes. In this paper, we introduce HOLM, Hallucinating Objects with Language Models, to address the challenge of partial observability. HOLM uses large pre-trained language models (LMs) to infer object hallucinations for the unobserved part of the environment. Our core intuition is that if a pair of objects co-appear in an environment frequently, our usage of language should reflect this fact about the world. Based on this intuition, we prompt language models to extract knowledge about object affinities which gives us a proxy for spatial relationships of objects. Our experiments show that HOLM performs better than the state-of-the-art approaches on two datasets for dRER; allowing to study generalization for both indoor and outdoor settings.

We propose a novel large-scale referring expression recognition dataset, Refer360°, consisting of 17,137 instruction sequences and ground-truth actions for completing these instructions in 360° scenes. Refer360° differs from existing related datasets in three ways. First, we propose a more realistic scenario where instructors and the followers have partial, yet dynamic, views of the scene – followers continuously modify their field-of-view (FoV) while interpreting instructions that specify a final target location. Second, instructions to find the target location consist of multiple steps for followers who will start at random FoVs. As a result, intermediate instructions are strongly grounded in object references, and followers must identify intermediate FoVs to find the final target location correctly. Third, the target locations are neither restricted to predefined objects nor chosen by annotators; instead, they are distributed randomly across scenes. This “point anywhere” approach leads to more linguistically complex instructions, as shown in our analyses. Our examination of the dataset shows that Refer360° manifests linguistically rich phenomena in a language grounding task that poses novel challenges for computational modeling of language, vision, and navigation.

We present an empirical analysis of state-of-the-art systems for referring expression recognition – the task of identifying the object in an image referred to by a natural language expression – with the goal of gaining insight into how these systems reason about language and vision. Surprisingly, we find strong evidence that even sophisticated and linguistically-motivated models for this task may ignore linguistic structure, instead relying on shallow correlations introduced by unintended biases in the data selection and annotation process. For example, we show that a system trained and tested on the input image without the input referring expression can achieve a precision of 71.2% in top-2 predictions. Furthermore, a system that predicts only the object category given the input can achieve a precision of 84.2% in top-2 predictions. These surprisingly positive results for what should be deficient prediction scenarios suggest that careful analysis of what our models are learning – and further, how our data is constructed – is critical as we seek to make substantive progress on grounded language tasks.

Learning syntactic categories is a fundamental task in language acquisition. Previous studies show that co-occurrence patterns of preceding and following words are essential to group words into categories. However, the neighboring words, or frames, are rarely repeated exactly in the data. This creates data sparsity and hampers learning for frame based models. In this work, we propose a paradigmatic representation of word context which uses probable substitutes instead of frames. Our experiments on child-directed speech show that models based on probable substitutes learn more accurate categories with fewer examples compared to models based on frames.