LibCAR Tier-1 LOO artifacts (MovieLens / FZU / Fudan): code and derived evaluation outputs

Derived Candidate-100 leave-one-out (LOO) scripts, aggregated result JSON, paired NDCG statistics, sparse-bin tables, library-domain baselines, MovieLens attribution outputs (including same-pool TALLRec-style LoRA), and optional Top-K checkpoints for the LibCAR manuscript (peer-review package). Source corpora are NOT rehosted. Obtain raw data from the Library Journal Data Center: FZU print circulation 2010-2021: hdl:20.500.12304/10842 Fudan 2017 Chinese e-journal access: hdl:20.500.12304/ERU2017 Headline checks in this package: FZU ?NDCG@10 ? +7.0% (CI [0.018, 0.019]); Fudan ?NDCG@10 ? +7.6% (CI [0.029, 0.031]) at flat HR@10. See README.md inside the zip. 2026-08-08 update: Fixed concept-DOI wording in README/SOURCE_DATA; added lora_tallrec_dense_paired5041_results.json (HR@10 0.169 / NDCG@10 0.089); set creator to named corresponding author for ESWA submission consistency. Legacy hybrid experiment.db / OpenClaw deposits remain superseded for offline LOO ranking evidence.