Self-alignment of Large Video Language Models with Refined Regularized Preference Optimization

Despite recent advances in Large Video Language Models (LVLMs), they still struggle with fine-grained temporal understanding, hallucinate, and often make simple mistakes on even simple video question-answering tasks, all of which pose significant challenges to their safe and reliable deployment in real-world applications. To address these limitations, we propose a self-alignment framework that enables LVLMs to learn from their own errors. Our proposed framework first obtains a training set of preferred and non-preferred response pairs, where non-preferred responses are generated by incorporating common error patterns that often occur due to inadequate spatio-temporal understanding, spurious correlations between co-occurring concepts, and over-reliance on linguistic cues while neglecting the vision modality, among others. To facilitate self-alignment of LVLMs with the constructed preferred and non-preferred response pairs, we introduce Refined Regularized Preference Optimization (RRPO), a novel preference optimization method that utilizes sub-sequence-level refined rewards and token-wise KL regularization to address the limitations of Direct Preference Optimization (DPO). We demonstrate that RRPO achieves more precise alignment and more stable training compared to DPO. Our experiments and analysis validate the effectiveness of our approach across diverse video tasks, including video hallucination, short- and long-video understanding, and fine-grained temporal reasoning.

Advancing Medical Representation Learning Through High-Quality Data

Despite the growing scale of medical Vision-Language datasets, the impact of dataset quality on model performance remains under-explored. We introduce Open-PMC, a high-quality medical dataset from PubMed Central, containing 2.2 million image-text pairs, enriched with image modality annotations, subfigures, and summarized in-text references. Notably, the in-text references provide richer medical context, extending beyond the abstract information typically found in captions. Through extensive experiments, we benchmark Open-PMC against larger datasets across retrieval and zero-shot classification tasks. Our results show that dataset quality-not just size-drives significant performance gains. We complement our benchmark with an in-depth analysis of feature representation. Our findings highlight the crucial role of data curation quality in advancing multimodal medical AI. We release Open-PMC, along with the trained models and our codebase.

A Shared Encoder Approach to Multimodal Representation Learning

Multimodal representation learning has demonstrated remarkable potential in enabling models to process and integrate diverse data modalities, such as text and images, for improved understanding and performance. While the medical domain can benefit significantly from this paradigm, the scarcity of paired multimodal data and reliance on proprietary or pretrained encoders pose significant challenges. In this work, we present a shared encoder framework for multimodal representation learning tailored to the medical domain. Our approach employs a single set of encoder parameters shared across modalities, augmented with learnable modality features. Empirical results demonstrate that our shared encoder idea achieves superior performance compared to separate modality-specific encoders, demonstrating improved generalization in data-constrained settings. Notably, the performance gains are more pronounced with fewer training examples, underscoring the efficiency of our shared encoder framework for real-world medical applications with limited data. Our code and experiment setup are available at https://github.com/VectorInstitute/shared_encoder.

Task-agnostic Prompt Compression with Context-aware Sentence Embedding and Reward-guided Task Descriptor

The rise of Large Language Models (LLMs) has led to significant interest in prompt compression, a technique aimed at reducing the length of input prompts while preserving critical information. However, the prominent approaches in prompt compression often require explicit questions or handcrafted templates for compression, limiting their generalizability. We propose Task-agnostic Prompt Compression (TPC), a novel framework that generalizes compression across tasks and domains without requiring input questions or templates. TPC generates a context-relevant task description using a task descriptor trained on a curated dataset of context and query pairs, and fine-tuned via reinforcement learning with a reward function designed to capture the most relevant information. The task descriptor is then utilized to compute the relevance of each sentence in the prompt to generate the compressed prompt. We introduce 3 model sizes (Base, Large, and Huge), where the largest model outperforms the existing state-of-the-art methods on LongBench and ZeroSCROLLS benchmarks, and our smallest model performs comparable to the existing solutions while being considerably smaller.

Scaling laws in wearable human activity recognition

Many deep architectures and self-supervised pre-training techniques have been proposed for human activity recognition (HAR) from wearable multimodal sensors. Scaling laws have the potential to help move towards more principled design by linking model capacity with pre-training data volume. Yet, scaling laws have not been established for HAR to the same extent as in language and vision. By conducting an exhaustive grid search on both amount of pre-training data and Transformer architectures, we establish the first known scaling laws for HAR. We show that pre-training loss scales with a power law relationship to amount of data and parameter count and that increasing the number of users in a dataset results in a steeper improvement in performance than increasing data per user, indicating that diversity of pre-training data is important, which contrasts to some previously reported findings in self-supervised HAR. We show that these scaling laws translate to downstream performance improvements on three HAR benchmark datasets of postures, modes of locomotion and activities of daily living: UCI HAR and WISDM Phone and WISDM Watch. Finally, we suggest some previously published works should be revisited in light of these scaling laws with more adequate model capacities.

SelfPrompt: Confidence-Aware Semi-Supervised Tuning for Robust Vision-Language Model Adaptation

We present SelfPrompt, a novel prompt-tuning approach for vision-language models (VLMs) in a semi-supervised learning setup. Existing methods for tuning VLMs in semi-supervised setups struggle with the negative impact of the miscalibrated VLMs on pseudo-labelling, and the accumulation of noisy pseudo-labels. SelfPrompt addresses these challenges by introducing a cluster-guided pseudo-labelling method that improves pseudo-label accuracy, and a confidence-aware semi-supervised learning module that maximizes the utilization of unlabelled data by combining supervised learning and weakly-supervised learning. Additionally, we investigate our method in an active semi-supervised learning setup, where the labelled set is strategically selected to ensure the best utilization of a limited labelling budget. To this end, we propose a weakly-supervised sampling technique that selects a diverse and representative labelled set, which can be seamlessly integrated into existing methods to enhance their performance. We conduct extensive evaluations across 13 datasets, significantly surpassing state-of-the-art performances with average improvements of 6.23% in standard semi-supervised learning, 6.25% in active semi-supervised learning, and 4.9% in base-to-novel generalization, using a 2-shot setup. Furthermore, SelfPrompt shows excellent generalization in single-shot settings, achieving an average improvement of 11.78%.

Dynamic Prototype Rehearsal for Continual Learning in ECG Arrhythmia Detection

Continual Learning (CL) methods aim to learn from a sequence of tasks while avoiding the challenge of forgetting previous knowledge. We present DREAM-CL, a novel CL method for ECG arrhythmia detection that introduces dynamic prototype rehearsal memory. DREAM-CL selects representative prototypes by clustering data based on learning behavior during each training session. Within each cluster, we apply a smooth sorting operation that ranks samples by training difficulty, compressing extreme values and removing outliers. The more challenging samples are then chosen as prototypes for the rehearsal memory, ensuring effective knowledge retention across sessions. We evaluate our method on time-incremental, class-incremental, and lead-incremental scenarios using two widely used ECG arrhythmia datasets, Chapman and PTB-XL. The results demonstrate that DREAM-CL outperforms the state-of-the-art in CL for ECG arrhythmia detection. Detailed ablation and sensitivity studies are performed to validate the different design choices of our method.

Federated Domain Generalization with Label Smoothing and Balanced Decentralized Training

In this paper, we propose a novel approach, Federated Domain Generalization with Label Smoothing and Balanced Decentralized Training (FedSB), to address the challenges of data heterogeneity within a federated learning framework. FedSB utilizes label smoothing at the client level to prevent overfitting to domain-specific features, thereby enhancing generalization capabilities across diverse domains when aggregating local models into a global model. Additionally, FedSB incorporates a decentralized budgeting mechanism which balances training among clients, which is shown to improve the performance of the aggregated global model. Extensive experiments on four commonly used multi-domain datasets, PACS, VLCS, OfficeHome, and TerraInc, demonstrate that FedSB outperforms competing methods, achieving state-of-the-art results on three out of four datasets, indicating the effectiveness of FedSB in addressing data heterogeneity.

Socially-Informed Reconstruction for Pedestrian Trajectory Forecasting

Pedestrian trajectory prediction remains a challenge for autonomous systems, particularly due to the intricate dynamics of social interactions. Accurate forecasting requires a comprehensive understanding not only of each pedestrian's previous trajectory but also of their interaction with the surrounding environment, an important part of which are other pedestrians moving dynamically in the scene. To learn effective socially-informed representations, we propose a model that uses a reconstructor alongside a conditional variational autoencoder-based trajectory forecasting module. This module generates pseudo-trajectories, which we use as augmentations throughout the training process. To further guide the model towards social awareness, we propose a novel social loss that aids in forecasting of more stable trajectories. We validate our approach through extensive experiments, demonstrating strong performances in comparison to state of-the-art methods on the ETH/UCY and SDD benchmarks.

CycleCrash: A Dataset of Bicycle Collision Videos for Collision Prediction and Analysis

Self-driving research often underrepresents cyclist collisions and safety. To address this, we present CycleCrash, a novel dataset consisting of 3,000 dashcam videos with 436,347 frames that capture cyclists in a range of critical situations, from collisions to safe interactions. This dataset enables 9 different cyclist collision prediction and classification tasks focusing on potentially hazardous conditions for cyclists and is annotated with collision-related, cyclist-related, and scene-related labels. Next, we propose VidNeXt, a novel method that leverages a ConvNeXt spatial encoder and a non-stationary transformer to capture the temporal dynamics of videos for the tasks defined in our dataset. To demonstrate the effectiveness of our method and create additional baselines on CycleCrash, we apply and compare 7 models along with a detailed ablation. We release the dataset and code at https://github.com/DeSinister/CycleCrash/ .