Artificial intelligence systems in critical fields like autonomous driving and medical imaging analysis often continually learn new tasks using a shared stream of input data. For instance, after learning to detect traffic signs, a model may later need to learn to classify traffic lights or different types of vehicles using the same camera feed. This scenario introduces a challenging setting we term Continual Multitask Learning (CMTL), where a model sequentially learns new tasks on an underlying data distribution without forgetting previously learned abilities. Existing continual learning methods often fail in this setting because they learn fragmented, task-specific features that interfere with one another. To address this, we introduce Learning with Preserving (LwP), a novel framework that shifts the focus from preserving task outputs to maintaining the geometric structure of the shared representation space. The core of LwP is a Dynamically Weighted Distance Preservation (DWDP) loss that prevents representation drift by regularizing the pairwise distances between latent data representations. This mechanism of preserving the underlying geometric structure allows the model to retain implicit knowledge and support diverse tasks without requiring a replay buffer, making it suitable for privacy-conscious applications. Extensive evaluations on time-series and image benchmarks show that LwP not only mitigates catastrophic forgetting but also consistently outperforms state-of-the-art baselines in CMTL tasks.

Forecasting companies’ financial metrics, such as profit or revenue, from textual data is typically heuristic and subjective due to the qualitative nature of business models and data. However, this paper shows that these metrics can be predicted with surprising accuracy using only a textual description of a company’s business and public data from peer companies using our novel framework CUBAN (Contextual Understanding of Business performance through Analysis of Neighboring companies). We introduce a multimodal transformer model with an annotation-gating mechanism that effectively integrates textual context with financial statements. Trained on 10-K reports from public companies since 2000, our model predicts the future log revenue and operating profit from descriptions of a company’s business and those of its peers (along with peer financial data), achieving a correlation coefficient (Pearson’s R) of 0.78 for log revenue prediction and a 79% F1-score for operating profit classification, demonstrating its efficacy in forecasting financial performance from primarily qualitative data.