Steven Yu (@lamsheeper) profile picture

Steven Yu

@lamsheeper
lamsheeper.github.io/
Joined October 2021
  • Tweets7
  • Following74
  • Followers76
Can we trace an AI model's answers back to its training data? Surprisingly, even when we know exactly which documents taught a model a fact, attribution methods often fail to find them. Our CATT benchmark investigates why and uncovers two simple fixes catt.baulab.info🧵
6
30
3
167
15,438
more replies
Does this generalize? On FTRACE-TREX, we remove each method's top-ranked training documents and retrain. If those documents really taught the facts, accuracy on those facts should drop. With our fixes, the accuracy drop is substantially larger for every method we test.
Image from Steven Yu (@lamsheeper)
1
1
1
11
748
Try the demo to see how each method ranks training documents! Paper: arxiv.org/abs/2610.40655 Demo: huggingface.co/spaces/catt-b… Code: github.com/Lamsheeper/catt Models & data: huggingface.co/catt-bench Joint work with @ShengYuWang6, @kimiahmdh, Antonio Torralba and @davidbau!
1
1
15
716
Load more