English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Language models (LMs) have been shown to memorize a great deal of factualknowledge contained in their training data. But when an LM generates anassertion, it is often difficult to determine where it learned this informationand whether it is true. In this paper, we propose the problem of fact tracing:identifying which training examples taught an LM to generate a particularfactual assertion. Prior work on training data attribution (TDA) may offereffective tools for identifying such examples, known as "proponents". Wepresent the first quantitative benchmark to evaluate this. We compare twopopular families of TDA methods -- gradient-based and embedding-based -- andfind that much headroom remains. For example, both methods have lowerproponent-retrieval precision than an information retrieval baseline (BM25)that does not have access to the LM at all. We identify key challenges that maybe necessary for further improvement such as overcoming the problem of gradientsaturation, and also show how several nuanced implementation details ofexisting neural TDA methods can significantly improve overall fact tracingperformance.

Towards Tracing Factual Knowledge in Language Models Back to the Training Data