OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training Tokens
Jiacheng Liu, Taylor Blanton, Yanai Elazar, Sewon Min, YenSung Chen, Arnavi Chheda-Kothary, Huy Tran, Byron Bischoff, Eric Marsh, Michael Schmitz, Cassidy Trier, Aaron Sarnat, Jenna James, Jon Borchardt, Bailey Kuehl, Evie Cheng, Karen Farley, Sruthi Sreeram, Taira Anderson, David Albright, Carissa Schoenick, Luca Soldaini
2025-04-10
Summary
This paper talks about OLMoTrace, a tool that helps track where AI language models get their answers by finding exact matches between their responses and the massive amount of text they were trained on.
What's the problem?
AI models often give answers without showing where the information came from, making it hard to know if they’re accurate or just making things up.
What's the solution?
OLMoTrace scans AI responses and links specific phrases directly to the original training documents, showing users the exact sources in seconds.
Why it matters?
This helps people trust AI answers more by letting them fact-check responses, spot made-up information, and understand how creative or original AI-generated content really is.
Abstract
We present OLMoTrace, the first system that traces the outputs of language models back to their full, multi-trillion-token training data in real time. OLMoTrace finds and shows verbatim matches between segments of language model output and documents in the training text corpora. Powered by an extended version of infini-gram (Liu et al., 2024), our system returns tracing results within a few seconds. OLMoTrace can help users understand the behavior of language models through the lens of their training data. We showcase how it can be used to explore fact checking, hallucination, and the creativity of language models. OLMoTrace is publicly available and fully open-source.