9 papers
Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
Praditha Alwis, Soumyadeep Chandra, Deepak Ravikumar +1
Reliable video understanding requires high-quality video datasets that can provide both precise semantic labels and temporally consistent annotations. Detecting annotation errors i…
The Easy Path to Robustness: Coreset Selection using Sample Hardness
Pranav Ramesh, Arjun Roy, Deepak Ravikumar +2
Designing adversarially robust models from a data-centric perspective requires understanding which input samples are most crucial for learning resilient features. While coreset sel…
TRIM: Token-wise Attention-Derived Saliency for Data-Efficient Instruction Tuning
Manish Nagaraj, Sakshi Choudhary, Utkarsh Saxena +2
Instruction tuning is essential for aligning large language models (LLMs) to downstream tasks and commonly relies on large, diverse corpora. However, small, high-quality subsets, k…
Coresets from Trajectories: Selecting Data via Correlation of Loss Differences
Manish Nagaraj, Deepak Ravikumar, Kaushik Roy
Deep learning models achieve state-of-the-art performance across domains but face scalability challenges in real-time or resource-constrained scenarios. To address this, we propose…
Towards Scalable Modeling of Compressed Videos for Efficient Action Recognition
Shristi Das Biswas, Efstathia Soufleri, Arani Roy +1
Training robust deep video representations has proven to be computationally challenging due to substantial decoding overheads, the enormous size of raw video streams, and their inh…
Finding the Muses: Identifying Coresets through Loss Trajectories
Manish Nagaraj, Deepak Ravikumar, Efstathia Soufleri +1
Deep learning models achieve state-of-the-art performance across domains but face scalability challenges in real-time or resource-constrained scenarios. To address this, we propose…
PIXELS: Progressive Image Xemplar-based Editing with Latent Surgery
Shristi Das Biswas, Matthew Shreve, Xuelu Li +2
Recent advancements in language-guided diffusion models for image editing are often bottle-necked by cumbersome prompt engineering to precisely articulate desired changes. An intui…
Curvature Clues: Decoding Deep Learning Privacy with Input Loss Curvature
Deepak Ravikumar, Efstathia Soufleri, Kaushik Roy
In this paper, we explore the properties of loss curvature with respect to input data in deep neural networks. Curvature of loss with respect to input (termed input loss curvature)…
Advancing Compressed Video Action Recognition through Progressive Knowledge Distillation
Efstathia Soufleri, Deepak Ravikumar, Kaushik Roy
Compressed video action recognition classifies video samples by leveraging the different modalities in compressed videos, namely motion vectors, residuals, and intra-frames. For th…