We investigate the use of in-context learning and prompt engineering to estimate the contributions of training data in the outputs of instruction-tuned large language models (LLMs). We propose two novel approaches: (1) a similarity-based approach that measures the difference between LLM outputs with and without provided context, and (2) a mixture distribution model approach that frames the problem of identifying contribution scores as a matrix factorization task. Our empirical comparison demonstrates that the mixture model approach is more robust to retrieval noise in in-context learning, providing a more reliable estimation of data contributions.
@article{arxiv.2408.11852,
title = {Fast Training Dataset Attribution via In-Context Learning},
author = {Milad Fotouhi and Mohammad Taha Bahadori and Oluwaseyi Feyisetan and Payman Arabshahi and David Heckerman},
journal= {arXiv preprint arXiv:2408.11852},
year = {2025}
}