过度保护的训练环境在测试时不足：让模型为自己的训练做出贡献
Overprotective Training Environments Fall Short at Testing Time: Let Models Contribute to Their Own Training
摘要：
尽管取得了重要进展，但会话系统通常会生成对人类来说听起来不自然的对话。我们推测原因在于他们不同的训练和测试条件：代理人在受控的“实验室”环境中接受训练，但在“野外”进行测试。在训练过程中，他们学习根据人类对话历史生成一个话语。另一方面，在测试过程中，它们必须相互作用，因此要处理噪声数据。我们建议通过使用包含人类和机器生成的对话样本的混合批次来训练模型来填补这一空白。我们评估了GuessWhat？！上提出的方法的有效性，一种视觉参照游戏。
英文摘要：
Despite important progress, conversational systems often generate dialogues that sound unnatural to humans. We conjecture that the reason lies in their different training and testing conditions: agents are trained in a controlled "lab" setting but tested in the "wild". During training, they learn to generate an utterance given the human dialogue history. On the other hand, during testing, they must interact with each other, and hence deal with noisy data. We propose to fill this gap by training the model with mixed batches containing both samples of human and machine-generated dialogues. We assess the validity of the proposed method on GuessWhat?!, a visual referential game.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
