In this paper, we propose reverse inference optimization (RIO), a simple and
effective method designed to enhance the robustness of
autoregressive-model-based zero-shot text-to-speech (TTS) systems using
reinforcement learning from human feedback (RLHF). To assess the quality of
speech produced by the TTS system without human annotations, RIO introduces a
novel concept termed as reverse inference based on the Bayesian principle,
which suggests that a high-quality generated speech should be able to be used
as a prompt for subsequent generation using the same TTS model. By leveraging
reverse inference as the standard to select exemplars used in RLHF from the
speech samples generated by the TTS system itself, RIO steers the subsequent
optimization towards a direction of enhancing the TTS robustness. The RIO
framework, comprising sampling, automatic annotating, and learning, obviates
the need for a reward model or pairwise preference data, and significantly
improves the stability of zero-shot TTS performance by reducing the
discrepancies between training and inference conditions. Our experimental
results verify that RIO can effectively improve both subjective and objective
metrics, including mean opinion scores, word error rates, and speaker
similarity. Remarkably, RIO can also diminish the incidence of bad outputs to
nearly zero percent, rivalling the robustness when using ground-truth speech as
the prompt.

逆向推断优化（RIO）是一种简单有效的方法，旨在使用来自人类反馈的强化学习，增强基于自回归模型的零样本文本到语音（TTS）系统的鲁棒性。RIO 通过引入基于贝叶斯原理的逆向推断的新概念来评估没有人类注释的 TTS 系统生成的语音质量，从而选择用于 RLHF 的示例，从而引导后续的优化以提高 TTS 的鲁棒性。RIO 框架通过采样、自动注释和学习来消除奖励模型或成对偏好数据的需求，并通过减少训练和推理条件之间的差异显著提高了零样本 TTS 性能的稳定性。实验结果验证了 RIO 能够有效改善主观和客观指标，包括平均意见分、词错误率和说话人相似性。值得注意的是，RIO 还可以将错误输出的发生率几乎降为零，与使用地面真实语音作为提示时的稳健性相媲美。

鲁棒零样本文本转语音合成与逆向推导优化

Robust Zero-Shot Text-to-Speech Synthesis with Reverse Inference  Optimization

Judiciously setting the base station transmit power that matches its
deployment environment is a key problem in ultra dense networks and
heterogeneous in-building cellular deployments. A unique characteristic of this
problem is the tradeoff between sufficient indoor coverage and limited outdoor
leakage, which has to be met without explicit knowledge of the environment. In
this paper, we address the small base station (SBS) transmit power assignment
problem based on stochastic bandit theory. Unlike existing solutions that rely
on heavy involvement of RF engineers surveying the target area, we take
advantage of the human user behavior with simple coverage feedback in the
network, and thus significantly reduce the planned human measurement. In
addition, the proposed power assignment algorithms follow the Bayesian
principle to utilize the available prior knowledge from system self
configuration. To guarantee good performance when the prior knowledge is
insufficient, we incorporate the performance correlation among similar power
values, and establish an algorithm that exploits the correlation structure to
recover majority of the degraded performance. Furthermore, we explicitly
consider power switching penalties in order to discourage frequent changes of
the transmit power, which cause varying coverage and uneven user experience.
Comprehensive system-level simulations are performed for both single and
multiple SBS deployment scenarios, and the resulting power settings are
compared to the state-of-the-art solutions. Significant performance gains of
the proposed algorithms are observed. Particularly, the correlation structure
enables the algorithm to converge much faster to the optimal long-term power
than other methods.

本文提出了一种基于随机赌博理论和贝叶斯原理的小型基站传输功率分配算法，其通过利用人类用户行为的简单覆盖反馈，极大地减少了计划人员的人力测量。该算法在考虑了功率切换惩罚的同时，将相似功率值之间的性能相关性纳入考虑，利用先前的先验知识恢复了更多的性能下降，并达到了最优长期功率设置。