PID Control-Based Self-Healing to Improve the Robustness of Large Language Models

Despite the effectiveness of deep neural networks in numerous natural language processing applications, recent findings have exposed the vulnerability of these language models when minor perturbations are introduced. While appearing semantically indistinguishable to humans, these perturbations can significantly reduce the performance of well-trained language models, raising concerns about the reliability of deploying them in safe-critical situations. In this work, we construct a computationally efficient self-healing process to correct undesired model behavior during online inference when perturbations are applied to input data. This is formulated as a trajectory optimization problem in which the internal states of the neural network layers are automatically corrected using a PID (Proportional-Integral-Derivative) control mechanism. The P controller targets immediate state adjustments, while the I and D controllers consider past states and future dynamical trends, respectively. We leverage the geometrical properties of the training data to design effective linear PID controllers. This approach reduces the computational cost to that of using just the P controller, instead of the full PID control. Further, we introduce an analytical method for approximating the optimal control solutions, enhancing the real-time inference capabilities of this controlled system. Moreover, we conduct a theoretical error analysis of the analytic solution in a simplified setting. The proposed PID control-based self-healing is a low cost framework that improves the robustness of pre-trained large language models, whether standard or robustly trained, against a wide range of perturbations. A detailed implementation can be found in:https://github.com/zhuotongchen/PID-Control-Based-Self-Healing-to-Improve-the-Robustness-of-Large-Language-Models.

Contrastive Learning with Consistent Representations

Contrastive learning demonstrates great promise for representation learning. Data augmentations play a critical role in contrastive learning by providing informative views of the data without needing the labels. However, the performance of the existing works heavily relies on the quality of the employed data augmentation (DA) functions, which are typically hand picked from a restricted set of choices. While exploiting a diverse set of data augmentations is appealing, the intricacies of DAs and representation learning may lead to performance degradation. To address this challenge and allow for a systemic use of large numbers of data augmentations, this paper proposes Contrastive Learning with Consistent Representations (CoCor). At the core of CoCor is a new consistency measure, DA consistency, which dictates the mapping of augmented input data to the representation space such that these instances are mapped to optimal locations in a way consistent to the intensity of the DA applied. Furthermore, a data-driven approach is proposed to learn the optimal mapping locations as a function of DA while maintaining a desired monotonic property with respect to DA intensity. The proposed techniques give rise to a semi-supervised learning framework based on bi-level optimization, achieving new state-of-the-art results for image recognition.