Finds the software, libraries and tools that scientific papers were built on, and who deserves credit for them.
GREP — the Great Research Extraction Project — reads scientific PDFs and extracts every software mention: the name, where it appears in the text, what kind of software it is, and what the authors did with it. Knowing what software science runs on is how the community gives credit to research software and understands its footprint.
Three independently trained models read every paper. Each was trained with a different recipe, so each makes different mistakes; a mention only enters the final record when at least two of the three find the same span, and the models that found it settle its type and purpose by majority. No single model's output becomes the answer on its own.
Underneath that, every individual work unit is run by three separate volunteers on the same model, and the head accepts it once two of them produce matching output. That defends the record against flaky hardware and bad actors without trusting any single machine.
Each leaf below runs the same science on different hardware or a different model. Pick whichever fits your machine.
Reads scientific PDFs and pulls out every software name, tool and library the authors used, with its location in the text, a type and a purpose. This leaf runs f14, the strongest of the three models o...
Reads scientific PDFs and pulls out every software name, tool and library the authors used, with its location in the text, a type and a purpose. This leaf runs f14, the strongest of the three models o...
Reads scientific PDFs and pulls out every software name, tool and library the authors used, with its location in the text, a type and a purpose. This leaf runs f13, trained with heavy weighting toward...
Reads scientific PDFs and pulls out every software name, tool and library the authors used, with its location in the text, a type and a purpose. This leaf runs f13, trained with heavy weighting toward...
Reads scientific PDFs and pulls out every software name, tool and library the authors used, with its location in the text, a type and a purpose. This leaf runs V1, the crowd's first-generation model. ...
Reads scientific PDFs and pulls out every software name, tool and library the authors used, with its location in the text, a type and a purpose. This leaf runs V1, the crowd's first-generation model. ...
