As large dialogue models become commonplace in practice, the problems
surrounding high compute requirements for training, inference and larger memory
footprint still persists. In this work, we present AUTODIAL, a multi-task
dialogue model that addresses the challenges of deploying dialogue model.
AUTODIAL utilizes parallel decoders to perform tasks such as dialogue act
prediction, domain prediction, intent prediction, and dialogue state tracking.
Using classification decoders over generative decoders allows AUTODIAL to
significantly reduce memory footprint and achieve faster inference times
compared to existing generative approach namely SimpleTOD. We demonstrate that
AUTODIAL provides 3-6x speedups during inference while having 11x fewer
parameters on three dialogue tasks compared to SimpleTOD. Our results show that
extending current dialogue models to have parallel decoders can be a viable
alternative for deploying them in resource-constrained environments.

AUTODIAL 是一种多任务对话模型，通过使用平行解码器来预测对话行为、领域、意图和状态，较之 SimpleTOD 等生成式方法，使用分类式解码器可大幅度减少内存占用并提高推理速度，在三种对话任务上拥有 11 倍少的参数以及 3-6 倍的推理速度提升，可为在资源有限的环境下部署对话模型提供可行的选择。

AUTODIAL：高效异步任务导向对话模型

AUTODIAL: Efficient Asynchronous Task-Oriented Dialogue Model

By inserting interleavers between intermediate stages of the polar encoder, a
new class of polar codes, termed interleaved polar (i-polar) codes, is
proposed. By the uniform interleaver assumption, we derive the weight
enumerating function (WEF) and input-output weight enumerating function (IOWEF)
averaged over the ensemble of i-polar codes. The average WEF can be used to
calculate the upper bound on the average block error rate (BLER) of a code
selected at random from the ensemble of i-polar codes. Also, we propose a
concatenated coding scheme that employs P high rate codes as the outer code and
Q i-polar codes as the inner code with an interleaver in between. The average
WEF of the concatenated code is derived based on the uniform interleaver
assumption. Simulation results show that BLER upper bounds can well predict
BLER performance levels of the concatenated codes. The results show that the
performance of the proposed concatenated code with P=Q=2 is better than that of
the CRC-aided i-polar code with P=Q=1 of the same length and code rate at high
signal-to-noise ratios (SNRs). Moreover, the proposed concatenated code allows
multiple decoders to operate in parallel, which can reduce the decoding latency
and hence is suitable for ultra-reliable low-latency communications (URLLC).

通过在极化编码器的中间阶段插入交织器，提出了一种新的极化码类别，称为交织极化码。本文利用均匀交织假设，得出了交织极化码集合上的加权枚举函数和输入输出加权枚举函数的平均值，同时提出了一种多码器联接编码方案，其中采用 P 个高速率码字作为外码，Q 个交织极化码字作为内码。实验结果表明，多码器联接编码的性能在高信噪比下比同样长度和码率下的 CRC 辅助交织极化码性能更好，同时适合于超可靠低延迟通信 (URLLC)。