SyncAI.news, a Varaisys broadcasting
Data Attribution Should Steer Pretraining
GM

Glenn Matlin, Chandreyi Chakraborty, Mohammed Rehan Parwani, Saehee Eom, Anthony Wen-Ming Zang, Akhil Theerthala, Arjun Chatterjee, Tejas Khandwekar, Brandon Michaels, Alvin Deng, Mark Riedl

· 1 min read

ResearcharXiv cs.LG

Data Attribution Should Steer Pretraining

arXiv:2611.19698v1 Announce Type: new Abstract: Pretraining data curation decides what a model learns from, yet what makes a document worth learning from depends on what the model already knows. Influence-driven selectors already rank documents to improve one run; we argue that training-data attribution should steer pretraining by identifying testable data properties that become reusable curation criteria, evaluated on new data against cheaper selectors at matched total cost, discovery included. A document ranking is consumed by the run it was computed for, whereas an explained data property can be tested across corpora and checkpoints, especially where familiar quality signals diverge from measured usefulness. Because historical contribution, an understandable explanation, and future data value require distinct evidence, we specify a feedback loop that interprets attributed sources, tests candidate data changes, and states for five claim types what evidence each requires and what none establishes. We close with a falsifiable comparison against perplexity, classifier, and mixing baselines at equal total compute.

Original source

This story was published by arXiv cs.LG and written by Glenn Matlin, Chandreyi Chakraborty, Mohammed Rehan Parwani, Saehee Eom, Anthony Wen-Ming Zang, Akhil Theerthala, Arjun Chatterjee, Tejas Khandwekar, Brandon Michaels, Alvin Deng, Mark Riedl. SyncAI.news shows a preview; the complete article is on the publisher's site.

Read the full story on arxiv.org

Similar News