Correct labels are indispensable for training effective machine learning models. However, creating high-quality labels is expensive, and even professionally labeled data contains errors and ambiguities. Filtering and denoising can be applied to curate labeled data prior to training, at the cost of additional processing and loss of information. An alternative is on-the-fly sample reweighting during the training process to decrease the negative impact of incorrect or ambiguous labels, but this typically requires clean seed data. In this work we propose unsupervised on-the-fly meta loss rescaling to reweight training samples. Crucially, we rely only on features provided by the model being trained, to learn a rescaling function in real time without knowledge of the true clean data distribution. We achieve this via a novel meta learning setup that samples validation data for the meta update directly from the noisy training corpus by employing the rescaling function being trained. Our proposed method consistently improves performance across various NLP tasks with minimal computational overhead. Further, we are among the first to attempt on-the-fly training data reweighting on the challenging task of dialogue modeling, where noisy and ambiguous labels are common. Our strategy is robust in the face of noisy and clean data, handles class imbalance, and prevents overfitting to noisy labels. Our self-taught loss rescaling improves as the model trains, showing the ability to keep learning from the model's own signals. As training progresses, the impact of correctly labeled data is scaled up, while the impact of wrongly labeled data is suppressed.

Deep neural networks are susceptible to backdoor attacks, where adversaries manipulate model predictions by inserting malicious samples into the training data. Currently, there is still a lack of direct filtering methods for identifying suspicious training data to unveil potential backdoor samples. In this paper, we propose a novel method, Prediction Shift Backdoor Detection (PSBD), leveraging an uncertainty-based approach requiring minimal unlabeled clean validation data. PSBD is motivated by an intriguing Prediction Shift (PS) phenomenon, where poisoned models' predictions on clean data often shift away from true labels towards certain other labels with dropout applied during inference, while backdoor samples exhibit less PS. We hypothesize PS results from neuron bias effect, making neurons favor features of certain classes. PSBD identifies backdoor training samples by computing the Prediction Shift Uncertainty (PSU), the variance in probability values when dropout layers are toggled on and off during model inference. Extensive experiments have been conducted to verify the effectiveness and efficiency of PSBD, which achieves state-of-the-art results among mainstream detection methods. Codes are available at https://github.com/WL-619/PSBD.

Due to the increasing need to handle the noisy label problem in a massive dataset, learning with noisy labels has received much attention in recent years. As a promising approach, there have been recent studies to select clean training data by finding small-loss instances before a deep neural network overfits the noisy-label data. However, it is challenging to prevent overfitting. In this paper, we propose a novel noisy-label detection algorithm by employing the property of overfitting on individual data points. To this end, we present two novel criteria that statistically measure how much each training sample abnormally affects the model and clean validation data. Using the criteria, our iterative algorithm removes noisy-label samples and retrains the model alternately until no further performance improvement is made. In experiments on multiple benchmark datasets, we demonstrate the validity of our algorithm and show that our algorithm outperforms the state-of-the-art methods when the exact noise rates are not given. Furthermore, we show that our method can not only be expanded to a real-world video dataset but also can be viewed as a regularization method to solve problems caused by overfitting.