• Publications
  • Influence
Deep Reinforcement Learning at the Edge of the Statistical Precipice
TLDR
This paper argues that reliable evaluation in the few-run deep RL regime cannot ignore the uncertainty in results without running the risk of slowing down progress in the field, and advocates for reporting interval estimates of aggregate performance and proposing performance profiles to account for the variability in results.
Data-Efficient Reinforcement Learning with Self-Predictive Representations
TLDR
The method, Self-Predictive Representations (SPR), trains an agent to predict its own latent state representations multiple steps into the future using an encoder which is an exponential moving average of the agent’s parameters and a learned transition model.
Pretraining Representations for Data-Efficient Reinforcement Learning
TLDR
This work uses unlabeled data to pretrain an encoder which is then finetuned on a small amount of task-specific data, and employs a combination of latent dynamics modelling and unsupervised goal-conditioned RL to encourage learning representations which capture diverse aspects of the underlying MDP.
GAIT: A Geometric Approach to Information Theory
TLDR
This work advocates the use of a notion of entropy that reflects the relative abundances of the symbols in an alphabet, as well as the similarities between them, and introduces geometry-aware counterparts for several concepts and theorems in information theory.
Crowdsourcing Text Simplification with Sentence Fusion
TLDR
A system for conducting text simplification using crowdsourcing, employing amateur human workers to simplify text at a sentence level is presented, and a graph-based sentence fusion system is introduced, which is used to augment the output of the human workers.
Improving Human Text Simplification with Sentence Fusion
TLDR
A graph-based sentence fusion approach to augment human simplification and a reranking approach to both select high quality simplifications and to allow for targeting simplifications with varying levels of simplicity are introduced.
Data-Efficient Reinforcement Learning with Momentum Predictive Representations
TLDR
This work trains an agent to predict its own latent state representations multiple steps into the future using an encoder which is an exponential moving average of the agent's parameters, and makes predictions using a learned transition model.
Human Evaluation for Text Simplification : The Simplicity-Adequacy Tradeoff
In this paper we examine human evaluation for text simplification. We find a strong inverse correlation between simplicity and adequacy, hinting that caution should be used when comparing systems
Iterated learning for emergent systematicity in VQA
TLDR
This work uses iterated learning to encourage the development of structure within an emergent language and shows that the resulting layouts support systematic generalization in neural agents solving the more complex task of visual question-answering.
...
1
2
...