Learning to Follow Language Instructions with Adversarial Reward Induction

By Dzmitry Bahdanau, Felix Hill, Jan Leike, Edward Hughes, Pushmeet Kohli, Edward Grefenstette. 2018

33 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Reinforcement Learning

Citations in the corpus

Scalable agent alignment via reward modeling: a research direction (2018, methodology): “Reinforcement learning from a learned reward function has been successful in gridworlds (Bahdanau et al., 2018), Atari games (Christiano et al., 2017; Ibarz et al., 2018), and continuous motor control tasks (Ho & Ermon, 2016; Christiano et al., 2017).”
BabyAI: A Platform to Study the Sample Efficiency of Grounded Language Learning (2018, background): “Secondly, BabyAI introduces partial observability (unlike the gridworld of Bahdanau et al. (2018)).”
Using Natural Language for Reward Shaping in Reinforcement Learning (2019, background): “More recently, [ Bahdanau et al. , 2018 ] proposed an adversarial learning framework wherein a discriminator distinguishes between a ﬁxed set of good (instruction, state) pairs and (instruction, state) pairs generated by the current policy, and this is used as a reward function to simultaneously…”
BabyAI: First Steps Towards Grounded Language Learning With a Human In the Loop (2018, background): “…Hermann et al. (2017), Chaplot et al. (2018) and Wu et al. (2018), in which the agent can only navigate environment but cannot, for example move things around, (2) partial observability (missing in the gridworld of Bahdanau et al. (2018)) and (3) a systematic definition of the synthetic language.”
Unsupervised Control Through Non-Parametric Discriminative Rewards (2018, unknown): “Similarly, AGILE (Bahdanau et al., 2018) learns an instruction-conditional policy where goals in a grid-world are speciﬁed in terms of predicates which should be satisﬁed, and a reward function is learned using a discriminator trained to distinguish states achieved by the policy from a dataset of…”
From Language to Goals: Inverse Reinforcement Learning for Vision-Based Instruction Following (2019, methodology): “Bahdanau et al. (2018); Tung et al. (2018) also learn language-conditioned reward functions, but do not perform IRL, meaning that the objective does not correspond to matching the expert’s trajectory distribution.”
RTFM: Generalising to Novel Environment Dynamics via Reading (2019, methodology): “We compare txt 2 π to the FiLM model by Bahdanau et al. (2019) and a language-conditioned residual CNN model.”