Replay to Remember: Retaining Domain Knowledge in Streaming Language Models

Continual learning in large language models (LLMs) typically encounters the critical challenge of catastrophic forgetting, where previously acquired knowledge deteriorates upon exposure to new data. While techniques like replay buffers and parameter-efficient tuning (e.g., Low-Rank Adaptation or LoRA) have been proposed, few studies investigate real-time domain adaptation under strict computational and data-stream constraints. In this paper, we demonstrate a lightweight method combining LoRA and a minimal replay mechanism in a realistic streaming setting across three diverse knowledge domains: medical question answering, genetics, and law. Using perplexity, semantic similarity, and GPT-based human-like evaluation metrics, we quantify the model's adaptation, forgetting, and recovery over time. Our experiments reveal that while catastrophic forgetting naturally occurs, even minimal replay significantly stabilizes and partially restores domain-specific knowledge. This study contributes practical insights for deploying adaptable LLMs in resource-constrained, real-world scenarios.

Graph-Augmented LSTM for Forecasting Sparse Anomalies in Graph-Structured Time Series

Detecting anomalies in time series data is a critical task across many domains. The challenge intensifies when anomalies are sparse and the data are multivariate with relational dependencies across sensors or nodes. Traditional univariate anomaly detectors struggle to capture such cross-node dependencies, particularly in sparse anomaly settings. To address this, we propose a graph-augmented time series forecasting approach that explicitly integrates the graph of relationships among time series into an LSTM forecasting model. This enables the model to detect rare anomalies that might otherwise go unnoticed in purely univariate approaches. We evaluate the approach on two benchmark datasets - the Yahoo Webscope S5 anomaly dataset and the METR-LA traffic sensor network - and compare the performance of the Graph-Augmented LSTM against LSTM-only, ARIMA, and Prophet baselines. Results demonstrate that the graph-augmented model achieves significantly higher precision and recall, improving F1-score by up to 10% over the best baseline

Variance-Aware Loss Scheduling for Multimodal Alignment in Low-Data Settings

Training vision-language models for image-text alignment typically requires large datasets to achieve robust performance. In low-data scenarios, standard contrastive learning can struggle to align modalities effectively due to overfitting and unstable training dynamics. In this paper, we propose a variance-aware loss scheduling approach that dynamically adjusts the weighting of the contrastive loss based on the statistical variability (uncertainty) in the model's alignment predictions. Using a subset of the Flickr8k image-caption dataset to simulate limited data conditions, we demonstrate that our approach improves image-text retrieval accuracy compared to a fixed-weight baseline. We also compare against other adaptive weighting strategies (using output entropy and cosine similarity spread) and find that variance-aware scheduling provides the best overall trade-off. Qualitatively, our method yields more distinct multimodal embeddings as shown by t-SNE visualizations. Moreover, in a stress test with noise-injected captions and images, the variance-guided loss proves more robust, maintaining higher recall when random perturbations are introduced. These results highlight the benefit of adaptive loss weighting for multimodal alignment in low-data regimes.