# Adaptive Bridge between Training and Inference for Dialogue

arXiv: https://arxiv.org/abs/2110.11560 · HTML: https://mlsanity.com/paper/2110.11560/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 31 references that are in the map.
- First posted on arXiv 22 Oct 2021.

## Builds on (most cited first, 23 of 23)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Sequence to Sequence Learning with Neural Networks](https://mlsanity.com/paper/1409.3215/) (2014)
- [Neural Machine Translation of Rare Words with Subword Units](https://mlsanity.com/paper/1508.07909/) (2015)
- [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019)
- [Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation](https://mlsanity.com/paper/1609.08144/) (2016)
- [A Diversity-Promoting Objective Function for Neural Conversation Models](https://mlsanity.com/paper/1510.03055/) (2015)
- [Show and Tell: A Neural Image Caption Generator](https://mlsanity.com/paper/1411.4555/) (2014)
- [Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks](https://mlsanity.com/paper/1506.03099/) (2015)
- [Sequence Level Training with Recurrent Neural Networks](https://mlsanity.com/paper/1511.06732/) (2015)
- [Neural Text Generation with Unlikelihood Training](https://mlsanity.com/paper/1908.04319/) (2019)
- [Long-term Recurrent Convolutional Networks for Visual Recognition and Description](https://mlsanity.com/paper/1411.4389/) (2014)
- [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://mlsanity.com/paper/2001.04063/) (2020)
- [Bridging the Gap between Training and Inference for Neural Machine Translation](https://mlsanity.com/paper/1906.02448/) (2019)
- [Minimum Risk Training for Neural Machine Translation](https://mlsanity.com/paper/1512.02433/) (2015)
- [Adversarial Learning for Neural Dialogue Generation](https://mlsanity.com/paper/1701.06547/) (2017)
- [TextGAIL: Generative Adversarial Imitation Learning for Text Generation](https://mlsanity.com/paper/2004.13796/) (2020)
- [Adversarial Neural Machine Translation](https://mlsanity.com/paper/1704.06933/) (2017)
- [ColdGANs: Taming Language GANs with Cautious Sampling Strategies](https://mlsanity.com/paper/2006.04643/) (2020)
- [Self-Adversarial Learning with Comparative Discrimination for Text Generation](https://mlsanity.com/paper/2001.11691/) (2020)
- [Greedy Search with Probabilistic N-gram Matching for Neural Machine Translation](https://mlsanity.com/paper/1809.03132/) (2018)
- [TeaForN: Teacher-Forcing with N-grams](https://mlsanity.com/paper/2010.03494/) (2020)

## Built on by (most cited first, 5 of 5)

- [Dynamically Retrieving Knowledge via Query Generation for Informative Dialogue Generation](https://mlsanity.com/paper/2208.00128/) (2022)
- [Promoting Open-domain Dialogue Generation through Learning Pattern Information between Contexts and Responses](https://mlsanity.com/paper/2309.02823/) (2023)
- [Towards Robust Online Dialogue Response Generation](https://mlsanity.com/paper/2203.03168/) (2022)
- [Bilevel Scheduled Sampling for Dialogue Generation](https://mlsanity.com/paper/2309.01953/) (2023)
- [UniRQR: A Unified Model for Retrieval Decision, Query, and Response Generation in Internet-Based Knowledge Dialogue Systems](https://mlsanity.com/paper/2401.06811/) (2024)

## Abstract (arXiv)

Although exposure bias has been widely studied in some NLP tasks, it faces its unique challenges in dialogue response generation, the representative one-to-various generation scenario. In real human dialogue, there are many appropriate responses for the same context, not only with different expressions, but also with different topics. Therefore, due to the much bigger gap between various ground-truth responses and the generated synthetic response, exposure bias is more challenging in dialogue generation task. What's more, as MLE encourages the model to only learn the common words among different ground-truth responses, but ignores the interesting and specific parts, exposure bias may further lead to the common response generation problem, such as "I don't know" and "HaHa?" In this paper, we propose a novel adaptive switching mechanism, which learns to automatically transit between ground-truth learning and generated learning regarding the word-level matching score, such as the cosine similarity. Experimental results on both Chinese STC dataset and English Reddit dataset, show that our adaptive method achieves a significant improvement in terms of metric-based evaluation and human evaluation, as compared with the state-of-the-art exposure bias approaches. Further analysis on NMT task also shows that our model can achieve a significant improvement.