Modularity Improves Out-of-Domain Instruction Following
By Rodolfo Corona, Daniel Fried, Coline Devin, Daniel Klein, Trevor Darrell. 2020
5 citations recorded by the source corpus.
We propose a modular architecture for following natural language instructions that describe sequences of diverse subgoals, such as navigating to landmarks or picking up objects. Standard, non-modular, architectures used in instruction following do not exploit subgoal compositionality and often struggle on out-of-distribution tasks and environments. In our approach, subgoal modules each carry out natural language instructions for a specific subgoal type. A sequence of modules to execute is chosen by learning to segment the instructions and predicting a subgoal type for each segment. When compared to standard sequence-to-sequence approaches on ALFRED, a challenging instruction following benchmark, we find that modularization improves generalization to environments unseen in training and to novel tasks.
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Oleksiy Ostapenko
- Massimo Caccia
- Pau Rodr'iguez L'opez
- Laurent Charlin
- Lucas Caccia
- Timothée Lesort
- Gyuhak Kim
- Quang Pham
Citations in the corpus
- Continual Learning via Local Module Composition (2021, background): “the number of tasks; (iii) systematic [8] and out-of-distribution (OOD) generalization [18] through knowledge recombination; and (iv) biological plausibility [91, 90, 96].”