vtok101-distr-attribution-baselines

vtok101 attribution baselines, with a hard negative beside every document

Data-attribution scores over lamsheeper-data-attribution/Qwen3.5-4B-d0-vtok101-distr-lora-seeds: 3 function counts x 7 document counts x 4 seeds, scored by 12 methods.

Each training document defines one synthetic constant function, and each query asks for one function's value. The ground truth for a query is the set of documents describing its function, so a method is measured by how far up its ranking those documents come.

Beside every real document the corpus carries a decoy: a shadow function <A07> taught the same constant in the same prompt template as the real <B07>, and nothing else. So the candidate pool is 2 x functions x documents and half of it is a near-twin of something. Queries only ever ask about a real function, so AUC means exactly what it means in the clean arm; Recall@R does not, its chance line being R/(2 x F x n) rather than R/(F x n).