AI RESEARCH
STRIDE: Training Data Attribution via Sparse Recovery from Subset Perturbations
arXiv CS.LG
•
Training Data Attribution (TDA) seeks to trace a model's predictions back to its training data. The gold standard for TDA relies on causal interventions, observing how a model changes when data is added or removed, but repeated retraining is computationally challenging for Large Language Models (LLMs). However, tracking gradients across billions of parameters is not only prohibitively expensive but reli