Work

Pre-registration: 1-round, 3-seed replication of the QED-gain claim for Conditional GraphGANFed (arXiv 2608.24610) on QM9

Year: 2026

Type: other

Abstract: Frozen pre-registration (13 Oct 2026) for a private replication test of one quantitative claim in arXiv 2608.24610, run with the authors' released code (cGraphGANFed d8299d40, GraphGANFed 85d245d6), deposited before the final runs. States the claim, reduced protocol (1 round, 3 seeds: 11-13), every patch to the released code, the evaluator, the decision rule and its planted-effect control, and what was seen beforehand. Includes the finding that the released code does not federate. Repository commit a0a48d04 of the AegisMind analysis tree. No results exist at deposit time.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.23327504

Author: Goodman, John

Topic: Privacy-Preserving Technologies in Data

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7221570698

Cites:

Cited by:

Related to: