The paper studies why language-model data-attribution methods fail, noting evaluations often vary several dataset properties at once. It introduces CATT, a controlled fact-tracing benchmark using fine-tuning on entirely fictitious knowledge with unambiguous ground truth. Why it matters Reliable...

Source: [Dev.to](https://dev.to/paragonaimag/controlled-benchmark-clarifies-failures-in-language-model-data-attribution-276)

Sponsored