Diffusion models have been successfully adapted to text generation tasks by mapping the discrete text into the continuous space. However, there exist nonnegligible gaps between training and inference, owing to the absence of the forward process during inference. Thus, the model only predicts based on the previously generated reverse noise rather than the noise computed by the forward process. Besides, the widely-used downsampling strategy in speeding up the inference will cause the mismatch of diffusion trajectories between training and inference. To understand and mitigate the above two types of training-inference discrepancies, we launch a thorough preliminary study. Based on our observations, we propose two simple yet effective methods to bridge the gaps mentioned above, named Distance Penalty and Adaptive Decay Sampling. Extensive experiments on \textbf{6} generation tasks confirm the superiority of our methods, which can achieve $100\times \rightarrow 200\times$ speedup with better performance.
The Pragmatic Persona: Discovering LLM Persona through Bridging Inference
Department of Artificial Intelligence, Chung-Ang University, Seoul, 06974, Republic of Korea（Chung-Ang大学人工智能系）；Department of Computer Science, University of British Columbia, Vancouver, BC V6T 1Z4, Canada（不列颠哥伦比亚大学计算机科学系）；Van Lang University, Ho Chi Minh City, Vietnam（文-lang大学）
Bridging the Gap between Training and Inference for Neural Machine Translation
Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）；University of Chinese Academy of Sciences（中国科学院大学）；Tencent Inc（腾讯公司）；Worcester Polytechnic Institute（伍斯特理工学院）；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）
Bridging the Gap Between Clean Data Training and Real-World Inference for Spoken Language Understanding
Peking University（北京大学）；The University of Sydney（悉尼大学）
口语语言理解（SLU）系统通常由多个流水线组件组成，其中每个组件严重依赖于其上游组件的结果。例如，意图检测（ID）和时隙填充（SF）需要其上游自动语音识别（ASR）将语音转换为文本。在这种情况下，上游扰动，例如ASR错误，环境噪声和粗心的用户说话，将传播到ID和SF模型，从而恶化系统性能。因此，预期性能良好的SF和ID模型在一定程度上是抗噪声的。然而，现有的模型是在干净的数据上训练的，这导致了\textit{gap between clean data training and real-world inference.}。为了弥补这一差距，我们从领域自适应的角度提出了一种方法，通过该方法，高质量和低质量的样本都嵌入到相似的向量空间中。同时，设计了去噪生成模型，降低了低质量样本的影响。在广泛使用的Snips数据集和大规模内部数据集（1000万个训练样本）上的实验表明，该方法不仅在真实世界（含噪）语料库上的性能优于基线模型，而且增强了鲁棒性，即在噪声环境下产生高质量的结果。源代码将被释放。
Bridging the Gap between Synthetic and Authentic Images for Multimodal Machine Translation
Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）；University of Chinese Academy of Sciences（中国科学院大学）；Beijing Language and Culture University（北京语言大学）
Bridging the Knowledge Void: Inference-time Acquisition of Unfamiliar Programming Languages for Coding Tasks
State Key Laboratory for Novel Software Technology, Nanjing University, China（新型软件技术国家重点实验室，南京大学，中国）；National Institute of Healthcare Data Science, Nanjing University, China（健康数据科学国家研究院，南京大学，中国）
