Generalization in Generation: A closer look at Exposure Bias

10/01/2019
by   Florian Schmidt, et al.
0

Exposure bias refers to the train-test discrepancy that seemingly arises when an autoregressive generative model uses only ground-truth contexts at training time but generated ones at test time. We separate the contributions of the model and the learning framework to clarify the debate on consequences and review proposed counter-measures. In this light, we argue that generalization is the underlying property to address and propose unconditional generation as its fundamental benchmark. Finally, we combine latent variable modeling with a recent formulation of exploration in reinforcement learning to obtain a rigorous handling of true and generated contexts. Results on language modeling and variational sentence auto-encoding confirm the model's generalization capability.

READ FULL TEXT
POST COMMENT

Comments

There are no comments yet.

Authors

page 1

page 2

page 3

page 4

05/25/2019

Quantifying Exposure Bias for Neural Language Generation

The exposure bias problem refers to the training-inference discrepancy c...
10/22/2021

Adaptive Bridge between Training and Inference for Dialogue

Although exposure bias has been widely studied in some NLP tasks, it fac...
02/28/2019

Evaluating Rewards for Question Generation Models

Recent approaches to question generation have used modifications to a Se...
06/29/2020

Learning Sparse Prototypes for Text Generation

Prototype-driven text generation uses non-parametric models that first c...
09/13/2021

Mitigating Catastrophic Forgetting in Scheduled Sampling with Elastic Weight Consolidation in Neural Machine Translation

Despite strong performance in many sequence-to-sequence tasks, autoregre...
09/04/2017

Learning to parse from a semantic objective: It works. Is it syntax?

Recent work on reinforcement learning and other gradient estimators for ...
07/27/2018

Auto-Encoding Variational Neural Machine Translation

We present a deep generative model of bilingual sentence pairs. The mode...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.